JP2006272013A - Radiation image processing equipment, image processing system, radiation image processing method, storage medium, and program - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide radiation image processing equipment and so on which creates corrective image data suitable for correcting the radiation image data of a subject to be photographed on the basis of sensitivity dispersion of two or more pixels comprising an image sensor from the image data obtained by performing radiation photography not via the subject to be photographed, but via an anti-scatter grid and performs the relevant correction using the relevant corrective image data. <P>SOLUTION: The radiation image processing equipment is comprised of a creating means which removes an image component resulting from the grid from the first image data obtained from radiation photography through the grid without the subject to be photographed and creates corrective image data and a correction means which corrects the second image data obtained by performing radiation photography via the subject to be photographed and the grid using the corrective image data. <P>COPYRIGHT: (C)2007,JPO&INPIT

Description

  The present invention relates to a radiographic image processing apparatus, an image processing system, a radiographic image processing method, a function of the apparatus or system, or a radiographic image processing apparatus that processes a radiographic image obtained by radiography using a grid for removing scattered radiation from a subject. The present invention relates to a program for causing a computer to implement or execute the processing steps of the method, and a computer-readable storage medium storing the program.

  In the case of observing a subject, particularly the inside of a human body, by detecting a transmission distribution of radiation, which is representative of X-rays that have passed through the subject, in recent years, a large-format image sensor using a semiconductor called a flat panel X-ray sensor ( It is becoming common to acquire a radiation image of a subject by directly receiving radiation transmitted through the subject by a solid-state imaging device).

  When the radiation passes through the subject, a part of the radiation is scattered, and the contrast of the image is lowered due to the influence. There has been no innovative progress in the technology for removing the scattered radiation (hereinafter also referred to as scattered radiation). Still, a member made of a plurality of lead plates arranged with a predetermined directivity called a scattered radiation removal grid (hereinafter also simply referred to as a grid) is disposed between the subject and the image sensor, and the scattered radiation removal grid is used to provide a predetermined Reduction of scattered radiation reaching the image sensor is performed by transmitting only radiation having directivity. However, when the scattered radiation removal grid is used, the shadow is reflected as striped image information (hereinafter also referred to as grid image) in the image acquired by the image sensor.

When an image sensor such as a flat panel X-ray sensor is used, the sensitivity of the plurality of pixels constituting the image sensor varies, and therefore image data acquired by the image sensor is corrected based on the sensitivity variation. Is done. For example, a flat panel X-ray sensor is manufactured by a semiconductor manufacturing technique and includes a plurality of pixels arranged in a plane. The plurality of pixels have variations in sensitivity due to the influence of the manufacturing accuracy in the semiconductor manufacturing process or the like, and thus the image correction as described above is necessary. In such image correction, generally, a reference image (also referred to as a correction image) is created based on an image obtained by photographing without a subject, and an image obtained by photographing the subject is used as the reference image. This is done by dividing pixel by pixel.
Japanese Patent Laid-Open No. 2000-3440

  When the above-described image correction is performed using the reference image created based on the radiographic image information obtained through the grid without passing through the subject, there are the following problems.

  First, since the position of the grid image in the image acquired by the image sensor changes due to the position change between the X-ray source and the image sensor, a stable reference image cannot be obtained.

  Secondly, the grid image in the radiographic image obtained by photographing through the subject and the grid changes in the influence of the subject and is different from the grid image in the reference image, so that a stable reference image can also be obtained. Can not.

  A change in the position of the grid image will be described with reference to FIG. In the figure, it is assumed that there is an X-ray source at a position indicated by 301a, a grid made of a plurality of lead foils exists at a position indicated by 303, and an image sensor is provided at a position indicated by 302. As shown in the figure, the shadow of a certain grid foil is shown at a position 304a. Next, in the figure, it is assumed that the position of the X-ray source has moved to a position 301b that is a distance D away from 301a. The shadow position of the same grid foil at that time moves to the position 304b. As a matter of fact, the distance (gap) T between the grid and the image sensor is not zero, and is about 5 mm due to protection of the image sensor, requirements for mechanical attachment, and the like. When the distance between the X-ray source and the grid is L, the movement distance ΔD of the shadow of the grid foil due to the movement of the X-ray source from 301a to 301b is represented by ΔD = DT / L. For example, if the distance L is 1800 mm, T is 5 mm, and D is 50 mm, ΔD≈0.14 mm. Since the distance between pixels of an image sensor that is usually used for diagnosis of a human body image is about 0.1 to 0.2 mm, the shadow of the grid foil on the image sensor is just a 5 cm movement of the X-ray source. Will move by a distance of about one pixel. In normal medical practice, it cannot be said that the position of the X-ray source does not move by about 5 cm between when the image is taken for obtaining the reference image and when the image is taken for obtaining the subject image. . In addition, the grid has a striped pattern in one direction (usually the vertical direction) (having a structure in which lead foils are arranged in a stripe shape when viewed from the X-ray incident direction), but the whole tilts due to the influence of mounting accuracy and the like ( There is also a possibility that the direction of the grid stripe is not stable with respect to the arrangement direction of the pixels of the image sensor.

  The contrast change of the grid image by the subject is a phenomenon in which the contrast of the grid image is lowered due to the influence of X-ray absorption, X-ray quality change or generation of scattered rays by the subject. Due to this phenomenon, a grid image having a different contrast from the grid image in the reference image is superimposed on the subject image.

  In such a case, even if the division of the subject image on which the grid image is superimposed and the reference image on which the grid image exists is performed, the grid image is emphasized on the contrary, rather than being erased or reduced. The grid image and the grid image in the reference image may be combined to give a complicated two-dimensional pattern.

  In addition, a technique for removing a periodic grid image existing on a subject image by a technique such as spatial filtering is known (see Patent Document 1). However, when a grid image is present in the reference image and the subject image is divided by the reference image as described above, a complex pattern is formed (superposed) on the corrected image without being removed. It is very difficult to remove the pattern from the corrected image by a technique such as spatial filtering.

  In addition, since there are no scattered rays in the case of shooting without passing through the subject, the reference image may be obtained by shooting without passing through the grid (the reference image should be obtained by shooting without passing through the grid). However, as a matter of fact, it is a complicated task to remove the grid every time it is taken and is not preferred.

  The present invention has been made to solve the above-described problems, and is used to correct radiographic image data of a subject based on sensitivity variations of a plurality of pixels constituting an image sensor (solid-state imaging device). A radiographic image processing apparatus that creates suitable image data for correction from image data obtained by radiography without passing through a subject and through a scattered radiation removal grid, and performs the correction using the image data for correction, an image It is an object to provide a processing system, a radiographic image processing method, a function of the apparatus or system or a program for causing a computer to realize or execute a processing step of the method, and a computer-readable storage medium storing the program To do.

  In order to achieve the above object, a first invention is a radiographic image processing apparatus for processing a radiographic image obtained by radiography using a grid for removing scattered radiation from a subject, A creation means for creating image data for correction by removing an image component caused by the grid from the first image data obtained by radiography through the grid, and a subject and the grid And correction means for correcting the second image data obtained by radiography using the correction image data.

  According to a second invention, in the first invention, the creation means creates the image component from the first image data based on the feature of the image component that is steady throughout the image. And removal means for removing the image component created by the image component creation means from the first image data.

  In a third aspect based on the first aspect, the creating means analyzes the first image data by analyzing at least the spatial frequency of the spatial frequency and direction of the periodic pattern exhibited by the image component. It has the analysis means to obtain.

  According to a fourth invention, in the third invention, the creation means extracts an extraction means for extracting a predetermined component including the image component from the first image data based on an analysis result of the analysis means, and the extraction The image processing apparatus includes processing means for processing the predetermined component obtained by the means to obtain the image component, and removal means for removing the image component obtained by the processing means from the first image data.

  In a fifth aspect based on the fourth aspect, the extracting means performs filtering for extracting, as the predetermined component, a component having the spatial frequency obtained by the analyzing means from the first image data. Features.

  In a sixth aspect based on the fourth aspect, the processing means estimates and processes an unsteady portion of the predetermined component from a steady portion of the predetermined component as the processing of the predetermined component. It is characterized by performing.

  In a sixth aspect based on the sixth aspect, the processing means detects the unsteady portion based on envelope information of the predetermined component.

  In an eighth aspect based on the sixth aspect, the processing means estimates an amplitude and a phase of a sine wave corresponding to the image component based on a stationary part of the predetermined component and the spatial frequency, The non-stationary part is repaired based on the estimation result.

  In a ninth aspect based on the eighth aspect, the processing means further performs filtering on the predetermined component after the repair to obtain the image component.

  In a tenth aspect based on the sixth aspect, the processing means substitutes the non-stationary part satisfying a predetermined condition among the non-stationary parts with a predetermined value to obtain the image component. And

  In an eleventh aspect based on the first aspect, the creation means executes the image component creation process for a predetermined line selected from the first image data.

  In a twelfth aspect based on the eleventh aspect, the creation means executes the creation process on an average result of a plurality of lines.

  In a thirteenth aspect based on the eleventh aspect, the creation unit obtains, as the image component of the line that has not been subjected to the creation process, the image acquired from the line that has been subjected to the creation process in the vicinity of the line. A component is used.

  In a fourteenth aspect based on the first aspect, the apparatus further comprises detection means for detecting mounting of the grid, and the creation means creates the correction image data based on a detection result of the detection means. Features.

  According to a fifteenth aspect, in the first aspect, the solid-state imaging device having an image receiving surface having a size larger than a range of the spatial distribution of the imaging target in the spatial distribution of the radiation intensity transmitted through the subject and the grid. An imaging means for acquiring a radiation image is provided.

  A sixteenth invention is characterized in that, in the second invention, image component storage means for storing the image component created by the image component creation means is provided.

  In a seventeenth aspect based on the first aspect, the creating means corrects the first image data using third image data obtained by radiography without passing through the subject and the grid. The image component creating means for creating the image component from the first image data, and the removing means for removing the image component created by the image component creating means from the first image data. Features.

  In an eighteenth aspect based on the seventeenth aspect, the first image component corrected by the third image data is based on a feature of the image component that the image component creation means is steady over the entire image. The image component is created from image data.

  A nineteenth aspect of the present invention is a radiographic image processing apparatus for processing a radiographic image obtained by radiography using a grid for removing scattered radiation from a subject, and the radiation does not pass through the subject and passes through the grid. An image component caused by the grid is removed from the first image data obtained by photographing, and a creation means for creating correction image data, and a second obtained by radiography through the subject and the grid Correction means for correcting the image data using the correction image data, and removal means for removing the image components caused by the grid from the second image data corrected by the correction means. Features.

  A twentieth invention is an image processing system configured such that a plurality of devices are communicably connected to each other, and at least one of the plurality of devices is any one of the first to nineteenth inventions. It has the function of a radiographic image processing apparatus.

  A twenty-first aspect of the present invention is a radiographic image processing method for processing a radiographic image obtained by radiography using a grid for removing scattered radiation from a subject, and the radiation does not pass through the subject and passes through the grid. A creation step of creating image data for correction by removing image components caused by the grid from the first image data obtained by photographing, and a second obtained by radiography through the subject and the grid And a correction step of correcting the image data using the correction image data.

  A twenty-second aspect of the present invention is a radiographic image processing method for processing a radiographic image obtained by radiography using a grid for removing scattered radiation from a subject, and the radiation does not pass through the subject and passes through the grid. A creation step of creating image data for correction by removing image components caused by the grid from the first image data obtained by photographing, and a second obtained by radiography through the subject and the grid A correction step for correcting the image data using the correction image data, and a removing unit for removing the image component due to the grid from the second image data corrected by the correction step. Features.

  A twenty-third aspect of the invention is a computer-readable recording medium storing a program for causing a computer to realize the function of the radiation image processing apparatus of any of the first to nineteenth aspects of the invention or the function of the image processing system of the twentieth aspect of the invention. Storage medium.

  A twenty-fourth invention is a computer-readable storage medium storing a program for causing a computer to execute the processing steps of the radiographic image processing method of the twenty-first or twenty-second invention.

  A twenty-fifth aspect of the invention is a program for causing a computer to realize the function of the radiation image processing apparatus of any of the first to nineteenth aspects of the invention or the function of the image processing system of the twentieth aspect of the invention.

  A twenty-sixth invention is a program for causing a computer to execute the processing steps of the radiographic image processing method of the twenty-first or twenty-second invention.

  As described above, according to the present invention, correction image data suitable for correcting subject image data based on variations in sensitivity of a plurality of pixels constituting an image sensor (solid-state imaging device) is not passed through the subject. A radiation image processing apparatus, an image processing system, a radiation image processing method, a function of the apparatus or system, or a processing step of the method that can be created from image data obtained by radiography through a scattered radiation removal grid A program for realizing or executing the program and a computer-readable storage medium storing the program can be provided.

[First Embodiment]
FIG. 1 schematically shows an X-ray imaging apparatus (radiation image processing apparatus) according to the first embodiment as a block diagram. In the figure, reference numeral 1 denotes an X-ray generator, which is controlled by a control device (not shown), is supplied with a high voltage, and emits X-rays in a direction indicated by an arrow. Reference numeral 2 denotes a subject represented by a human body, and reference numeral 3 denotes a bed supporting the subject 2. Reference numeral 4 denotes a scattered radiation removing grid arranged for the purpose of removing scattered radiation emitted from the subject and mainly selectively transmitting direct radiation. Reference numeral 5 denotes an X-ray image sensor (flat panel sensor) used for the purpose of converting the intensity distribution (X-ray image) of X-rays transmitted through the subject into an electric signal, and a plurality of X-ray image sensors (flat panel sensors) arranged in a planar shape and a matrix shape This is constructed using a large solid-state image sensor composed of the pixels. Hereinafter, this X-ray image sensor is referred to as a flat panel sensor. The X-ray image is spatially sampled on a two-dimensional plane by the flat panel sensor 5. When photographing a normal human body internal structure (human body part), the sampling pitch is set to about 100 to 200 μm. The flat panel sensor 5 is controlled by a controller (not shown), and the electric charges generated for each pixel in accordance with the incident X-ray dose are sequentially scanned and converted into an electric quantity (voltage or current). Output image data as an electrical signal.

  Reference numeral 6 denotes an A / D converter that converts an analog electric quantity output from the flat panel sensor 5 into a digital value. Reference numeral 7 denotes a memory (storage unit) that temporarily stores A / D converted digital values as image information. Reference numeral 8 denotes switching means for reading the memory 7 and selectively storing the read image information in the two memories (storage units) 9 or 10. The memory 9 is a memory that captures an image without emitting X-rays and stores an image signal output from the flat panel sensor as an offset fixed pattern image. The memory 10 is an object 2 obtained by exposing the X-rays. Is a memory for storing the images.

  For specific imaging, an X-ray dose measuring device (not shown) (usually called a phototimer) that monitors the X-ray dose that has passed through the subject is used for X-ray exposure control, and the integrated value of the exposed X-ray dose is a predetermined value. The X-ray exposure is stopped at the moment of becoming. The controller of this apparatus scans the flat panel sensor 5 immediately after the X-ray exposure stops, fetches the subject image information into the memory 7, sets the switching means 8 to the A side, and stores the image information in the memory 7 into the memory. 10 is stored. Immediately thereafter, the flat panel sensor 5 is driven without performing X-ray exposure, and charges are accumulated for the same time as the imaging time (X-ray exposure time) determined using the phototimer described above. Thereafter, the image data output by scanning the flat panel sensor 5 is stored in the memory 7 as offset fixed pattern data. The offset fixed pattern image data is stored in the memory 9 by setting the switching means 8 to the B side.

  Reference numeral 11 denotes a difference calculator that performs a difference calculation between two images stored in the memories 9 and 10, and subtracts each pixel value in the memory 9 corresponding to the position from each pixel value in the memory 10 in effect. Performs computation. Reference numeral 12 denotes a memory (storage unit) that stores the calculation result of the difference calculator 11.

  A reference table (Look Up Table, LUT) 13 converts image data stored in the memory 12 into logarithmic values in order to perform division by subtraction. Reference numeral 14 denotes switching means for selectively storing data from the LUT 13 in two memories (storage units) 15 or 16. The subject image data stored in the memory 12 is logarithmically converted by the LUT 13 and stored in the memory 15 by setting the switching means 14 to the C side.

  The memory 16 is a memory for storing an image obtained when an operation called calibration imaging is performed in the X-ray imaging apparatus. In the calibration shooting, image data is acquired by the same operation as described above, and the image data is stored in the memory 16 by setting the switching unit 14 to the D side, but the grid 4 is not connected to the subject 2. This is different from normal subject shooting.

  Normally, this calibration shooting is performed about once a day, for example, at the start of work, etc., and by this operation, image correction based on sensitivity variations (also referred to as gain variations) of a plurality of pixels constituting the flat panel sensor is performed. Reference image data (also referred to as correction image data) to be obtained is acquired.

  Reference numeral 17 denotes a grid image removing unit (selectively removing only the striped image component caused by the grid from the image data obtained by photographing through the grid 4 without passing through the subject 2 and stored in the memory 16. Grid image removal unit). Reference numeral 18 denotes a memory (storage unit) that stores reference image data (correction image data) corresponding to the processing result of the grid image removing unit 17, that is, sensitivity variation (gain variation) data of a plurality of pixels constituting the flat panel sensor. is there.

  Reference numeral 19 denotes a subtracting means (subtracting unit), which subtracts the reference image data stored in the memory 18 from the image data of the subject stored in the memory 15 (since each image data has undergone a logarithmic conversion LUT 13, Is equivalent to division between images). Reference numeral 20 denotes a memory (storage unit) that stores image data after image correction is performed based on sensitivity variations (gain variations) of a plurality of pixels constituting the flat panel sensor by the subtracting unit 19.

  FIG. 2 shows a configuration of the grid image removing unit 17. Reference numeral 171 denotes a memory (storage unit) for temporarily storing image data stored in the memory 16. Reference numeral 173 denotes a grid from the image data stored in the memory 171. Grid component extraction means (grid component extraction unit) for extracting only the striped image component resulting from the grid based on the feature of the resulting striped image component, 174 is caused by the grid extracted by the grid component extracting means 173 A memory (storage unit) for storing the striped image component data, 172 is a subtracting unit (subtraction unit), which subtracts the image data stored in the memory 174 from the image data stored in the memory 171. Standard image data (correction image data) is obtained.

  Although details will be described later, the grid component extraction means 173 basically uses the property that grid stripes (image components resulting from the grid) have a constant spatial frequency and are steady throughout the entire image. Perform the process. That is, a grid stripe component is extracted from the target image by spatial filtering, an unsteady portion is found from the grid stripe component, and the original grid stripe component is created by converting the unsteady portion into steady data. .

  Since the reference image data stored in the memory 18 of FIG. 1 does not include image components due to the grid at the time of calibration shooting, the image data of the subject stored in the memory 20 includes the grid at the time of shooting the subject. Only the image component resulting from is superimposed. Accordingly, reference numeral 21 in FIG. 1 denotes a grid image removing unit similar to the grid image removing unit 17, but the grid image removing unit 21 converts the grid stripe component (image component resulting from the grid) superimposed at the time of subject photographing into a grid image. It can be removed by the same action as the removing means 17.

  The image data after being processed by the grid image removing means 21 may be subjected to further processing, for example, suitable for use in image diagnosis such as gradation processing, dynamic range change processing, or spatial frequency processing. After the image processing is performed, the image data is transferred to an external device typified by a display device, a printing device, a filing device, or the like.

  In this embodiment and other embodiments described below, the grid stripe component may not be completely removed, for example, the grid stripe component may remain somewhat. However, even in such a case, if the grid stripe component is sufficiently reduced, the effect of each embodiment can be obtained, and the object of each embodiment can be achieved.

[Second Embodiment]
FIG. 3 is a block diagram illustrating an X-ray imaging apparatus (radiation image processing apparatus) according to the second embodiment. Components that operate in the same manner as the X-ray imaging apparatus of FIG. The description is omitted as appropriate.

  The feature of the second embodiment is that a memory (storage unit) 22 and a difference calculator 23 are used. The memory 22 stores sensitivity variation data (gain) of the flat panel sensor obtained by imaging the X-rays exposed without passing through the subject 2 and the grid 4 when the X-ray imaging apparatus is manufactured or installed. Variation data) is held. The sensitivity variation data is logarithmically converted and can be acquired by using the same components 1, 3 and 5 to 16 as those of the X-ray imaging apparatus of FIG. The data can be stored in the memory 22. The difference calculator 23 is a sensitivity stored in the memory 22 from the image data stored in the memory 16 obtained by mounting the grid and shooting without a subject in the calibration shooting described in the first embodiment. The variation data is subtracted (this subtraction substantially corresponds to division between images since both image data are logarithmically transformed). By this subtraction process, only the grid stripe component is extracted, and the obtained grid stripe component data is stored in the memory 18. Here, the image data stored in the memory 16 is image data obtained by detecting the intensity distribution of the X-ray modulated by the grid by the flat panel sensor, that is, sensitivity variation data and grid stripe component data of the flat panel sensor. Are the superimposed image data, and by subtracting (substantially dividing) the sensitivity variation data stored in the memory 22 from the image data stored in the memory 16, only the grid stripe component is extracted. It will be. The sensitivity variation data stored in the memory 22 needs to be substantially equal to the component of the sensitivity variation data of the flat panel sensor in the image data stored in the memory 16, and thus is stored in the memory 22 and the memory 16. It is necessary to make the photographing conditions substantially equal when obtaining both image data.

  However, as an actual problem, there is an imaging environment or an X-ray irradiation distribution between the time of obtaining image data stored in the memory 22 such as at the time of manufacture or installation of the X-ray imaging apparatus and the time of calibration imaging. Due to the difference, it may be difficult to extract only the grid stripe component as described above. In such a case, by further providing a grid component extraction unit 24 similar to the grid component extraction unit 173 that extracts only the grid stripe component from the image data obtained by the difference calculator 23, the grid stripe can be accurately detected. Ingredients can be extracted. A difference calculator 25 subtracts the image data obtained by the difference calculator 23 or the grid component extraction means 24 from the image data stored in the memory 16. The difference calculator 25 obtains only image data of sensitivity variation data (gain variation data) of a plurality of pixels of the flat panel sensor, and the image data is stored in the memory 18 as reference image data (correction image data). .

  The gist of the second embodiment is to improve the extraction accuracy of the grid stripe components. The image data stored in the memory 16 is obtained by adding the grid stripe component and the sensitivity variation component of a plurality of pixels of the flat panel sensor. The variation in sensitivity of a plurality of pixels of the flat panel sensor is not stationary, and its spatial change is very steep or random. For this reason, it is not easy to separate the sensitivity variation component and the grid stripe component. Accordingly, as in the present embodiment, image data resulting from only the sensitivity variation (gain variation) of the flat panel sensor acquired and stored at an appropriate time such as when the X-ray imaging apparatus is manufactured or installed is used. Thus, by removing or reducing unsteady image components resulting from the sensitivity variation, it is possible to improve the grid stripe component extraction accuracy.

  In the second embodiment, if the grid stripe component can be extracted with sufficient accuracy only by the processing of the difference calculator 23, the grid component extraction means 24 can be omitted.

[Third to Seventh Embodiments for Grid Component Extraction]
Hereinafter, third to seventh embodiments for illustrating a detailed configuration example of the grid image removing unit 21 in the first and second embodiments will be described with reference to the drawings. Note that the grid image removing unit 17 in the first embodiment (the same applies to the grid component extracting unit 173 in the first embodiment and the grid component extracting unit 24 in the second embodiment) as described above. This is a component that performs the same operation as that of the grid image removing unit 21 in the first and second embodiments, and the processing target is not an X-ray image that passes through the subject and the grid, but does not pass through the subject and that passes through the grid. Since it only becomes a line image (calibration photographed image) and can be configured in the same manner as the grid image removing unit 21, it is configured in the same manner as the configuration example in the following embodiment.

[Outline of Third to Seventh Embodiments]
Here, as embodiments of the present invention, the third to seventh embodiments are exemplified. Prior to specific description of the third to seventh embodiments, an outline of these will be described.

  Here, X-rays are used as an example of radiation, and an X-ray image obtained by X-ray imaging is processed.

  The configuration described below is an example configuration to which the present invention is applied, and is not limited thereto.

  First, the present invention solves the problem of removing grid stripes by the conventional filtering process (simple filtering process) as described in JP-A-3-12785, etc., by the following configuration. did.

  In other words, grid stripes that are superimposed on a signal of an X-ray image to be processed (hereinafter also simply referred to as “target image signal” or “target image”) and should exist as a stable stripe pattern over the entire image. Component information (hereinafter, also simply referred to as “grid stripe component”). Then, the grid stripe component is removed from the target image. For example, when the target image signal is an image signal after logarithmic conversion, the obtained grid stripe information is subtracted from the target image signal. This makes it possible to remove a stable grid stripe component without affecting the target image signal.

  Specifically, for example, based on the spatial frequency indicated by the grid stripe component, a component that substantially includes the grid stripe component is separated, and the component after separation is processed based on feature information that the grid stripe will indicate. The processed information is regarded as a grid stripe component and is removed from the target image signal.

  The grid fringe component has a fairly strong component according to the spatial spectrum expression. If the spatial frequency of the grid stripe is appropriately selected, the Nyquist frequency at the time of sampling (spatial frequency corresponding to 1/2 of the sampling frequency) It can exist in the vicinity. As a result, a state in which the grid stripe component does not overlap with the main component of the normal image signal as shown in FIGS. 25A to 25D can be easily obtained.

  Here, only when a steep fluctuation component (edge portion or the like) exists in the target image signal, only the grid stripe component is separated from the target image signal as shown in FIGS. 26 (a) to (d). Difficult to do.

  In some cases, the target image may include a region where the grid stripes themselves do not exist. This is because X-ray imaging of a subject including a part that almost completely blocks X-rays, or strong X-rays more than specified by the dynamic range of a flat panel sensor (hereinafter simply referred to as a sensor) When reaching a partial area of the sensor, the grid stripe component of the area disappears due to saturation.

  Normally, when an X-ray image is acquired, since the X-ray dose transmitted through the subject is emphasized, the X-ray dose is several hundred times as large as the subject transmission amount outside the subject (X-ray element missing region). In general, it is meaningless to expand the dynamic range of a sensor or sensor amplifier to the outside of the subject without information, and in most cases, the input / output characteristics of the sensor exhibit nonlinearity due to saturation. In the area to be presented, there is no grid stripe component or the contrast is lowered.

  Therefore, in the X-ray imaging apparatus of the following embodiment, when the target image data includes a steep fluctuation portion (for example, an edge portion) that makes it difficult to separate only the grid component from the target image signal, Alternatively, it is configured to adaptively cope with a case where a region where the grid stripe itself does not exist is included, and to remove only the grid stripe component without generating an artifact.

  The outline of the third to seventh embodiments as specific examples will be described below.

  In the third embodiment, one frame of a target image (an image obtained by X-ray imaging) is analyzed. Then, for a pixel defect in a direction orthogonal to the grid stripe, a linear prediction type pixel defect correction process is performed.

  For the grid stripe component, based on the analysis result, the grid stripe component is once extracted by low-order FIR filtering, and then the filtering result and the result of performing another FIR filtering on the filtering result The envelope information of the former filtering result is acquired by calculating the vector amplitude of. Then, based on the envelope information, the unsteady portion is extracted from the grid stripe component, and the unsteady portion is repaired using the surrounding steady portion, so that the grid stripe component as a whole Obtained as a steady signal sequence. Further, in order to more appropriately extract only the grid stripe component, a filtering process for selectively extracting an image component having a spatial frequency corresponding to the grid stripe component is performed on the stationary signal sequence, and the result is obtained. The grid stripe component.

  The grid stripe component acquired as described above is subtracted from the target image signal. Thereby, an image after removal of the grid stripe component is obtained.

  In the fourth embodiment, the target image is analyzed, and based on the result, pixel defect correction is performed on the image after removing the grid stripe component. As the pixel defect correction here, for example, pixel defect correction based on the average of surrounding pixel values can be applied.

  In the fifth embodiment, a detection unit that detects the mounting of the grid is provided, and when the grid mounting is confirmed by the detection unit, the target image is analyzed, and based on this result, predictive pixel defect correction, In addition, the grid stripe component is removed.

  In the sixth embodiment, when the X-ray irradiation field is in a state of being narrowed down so as to correspond to the imaging target region of the subject, only the partial image corresponding to the irradiation field is set as the target image. The process in the form of is executed.

  In the seventh embodiment, when there is no grid stripe in a partial region of the target image, a method of not performing the grid stripe removal process on the image data of a portion where no grid stripe exists.

  Specifically, in the generated grid stripe component data, the data of the portion (area) where the grid stripe component does not exist is replaced with zero (“0”) data. Do not do.

  In the third to seventh embodiments, for example, since the extracted grid stripe component is image information, the image may be held. Thereby, even when the grid stripe component is removed from the target image, the original target image, that is, the image before removing the grid stripe, is used by using the stored image information of the grid stripe component. May be recoverable.

Hereinafter, the grid stripe component removal process in the third to seventh embodiments will be described in detail.
In the following description, the third to seventh embodiments are collectively referred to as “this embodiment”.

The grid stripe component removal processing is mainly processing including the following first processing step to third processing step.
First processing step:
Line data in a direction orthogonal to the grid stripe is sampled from the target image obtained by including the grid stripe component, and the spatial frequency of the grid stripe is determined.
Second processing step:
Sequentially extract the grid fringe component from the target image and consider the generation of artifacts when subtracting the result (grid fringe component) from the target image. Image components mainly including grid stripes are extracted by FIR filtering.
Third processing step:
Based on the spatial frequency of the grid stripe obtained in the first processing step, the envelope of the image component mainly including the grid stripe obtained in the second processing step is converted into the phase of the component by FIR filtering different from that component. Is obtained by performing vector amplitude calculation with the component moved 90 °.

The grid stripe component removal processing including the first processing step to the third processing step will be described more specifically. First, the envelope information obtained in the third processing step always takes a positive value, The features include the following features (1) and (2).
(1) A very large value is shown for a steep fluctuation portion (for example, an edge portion). (2) For a portion where there is no grid stripe, a small value that is substantially “0” is shown.

  In the present embodiment, based on the above-described envelope information, a value portion (very large value portion) indicated by the feature (1) and a value portion (very small value) indicated by the feature (2). By repairing the grid stripe component, the creation of a more stable grid stripe component is realized.

  As a method of repairing the grid stripe component, for example, there is a method of replacing the portion of the feature (1) with a component predicted from a stable grid stripe portion in the vicinity thereof to obtain a stable grid stripe component as a whole. Can be mentioned.

  A normal filtering process is performed on the entire line with respect to a component mainly composed of grid stripes obtained only as described above and including only a stable component. At this time, filtering is performed to extract only image components having a narrower range of spatial frequencies centered on the spatial frequency of the grid stripes.

  The filtering result (extracted component) is the grid stripe component in the target image. When the envelope information has the characteristic (2), that is, the grid stripe component is mainly included in the grid stripe component. When there is a non-existing portion, the grid stripe component of the portion does not originally exist in that portion, so the grid stripe component of the portion is replaced with “0”.

  In addition, when performing a filtering process on a target image, a fast Fourier transform algorithm may be used in order to perform a steep filtering process stably and at high speed. In this case, the number of data points is “2” to the nth power. (N is a positive integer). For this reason, "0" is padded around the normal data so as to match the points. The data of the portion (area) filled with “0” may be considered as data corresponding to the portion whose envelope information indicates the feature (2).

  The effective spatial frequency as the spatial frequency of the grid itself is 30% to 40% of the sampling frequency of the sensor (reciprocal of the spatial sampling pitch) proposed in Japanese Patent Application No. 2000-028161 and the like. It is effective to be selected from a spatial frequency (60% to 80% of the Nyquist frequency). This is because the main component of the image is generally concentrated at 30% or less of the sampling frequency, and the component of strong grid stripes having a spatial frequency of 40% or more and 60% or less of the sampling frequency is an interpolation similar to linear interpolation after sampling. This is because when the processing is performed, it seems that another periodic amplitude fluctuation occurs, and the grid stripe itself is not stable.

  When the spatial frequency of the grid itself is “fg [cyc / mm]” and the sampling pitch of the sensor is “T”, the spatial frequency fm of the grid stripe is

なる式（１）で表される。

It is represented by the following formula (1).

  In the present embodiment, in consideration of the fact that a striped pattern corresponding to the spatial frequency fm represented by the above formula (1) exists in the target image as a grid striped component, the grid striped in the first processing step described above. Is extracted accurately. In other words, since the spatial frequency fg of the grid stripe is known in advance, the vicinity of the spatial frequency fm of the target image is searched from the sampled line data, and the spatial frequency indicating the peak value is obtained from the search result. Thus, the spatial frequency fm of the grid stripe in the target image is considered.

  Further, in the second processing step described above, the FIR filtering of the smallest possible span with the spatial frequency fm as the center is performed on the spatial frequency fm, so that the effective image components are almost removed and the steepness is sharp. The grid fringe component is roughly extracted in a state where the influence of the artifact due to the fluctuation portion (for example, the edge portion) is within a narrow range.

  At this time, in order to eliminate the phase fluctuation, for example, the coefficient series of the FIR filter is an even function, and a three-point or five-point FIR filter is desirable to satisfy the above narrow range.

  Specifically, for example, when a FIR filter having three symmetrical points is used and the coefficients are (a1, b1, a1), the response at the spatial frequency fm is “1” in order to obtain the coefficients (a1, b1, a1). Two conditions can be used: a condition of “0” and a condition that the DC component, which is the center value of the image information, is set to “0”.

That is, the coefficient calculation is
2 * a1 + b1 = 0
2 * a1 * cos (2πfmT) + b1 = 1
It is a simultaneous equation represented by the formula

なる式（２）で表される解をとる。

The solution represented by the following formula (2) is taken.

  In the FIR filtering described above, the response is “1” at the spatial frequency fm, but the response gradually increases when the spatial frequency becomes higher. In general, there is no image component in this portion, so that the grid stripes can be sufficiently extracted even with the FIR filtering.

  In addition, when FIR filtering is performed with five symmetrical points and the coefficients are (a2, b2, c2, b2, a2), the coefficients (a2, b2, c2, b2, a2) are obtained at the spatial frequency fm. In addition to the two conditions of the condition that the response is “1” and the condition that the DC component that is the center value of the image information is “0”, the differential value of the response at the spatial frequency fm is “0” (peak ) Can also be used.

That is, the coefficient calculation is performed by performing a simple calculation from the solution represented by the above formula (2).
(-A1 ² , 2a1 (1-b1), 1-2a1 ^2- (1-b1) ² , 2a1 (1-b1), -a1 ² )
The following solution is obtained.

  As a method of obtaining a filter for symmetric five-point FIR filtering, for example, first, when a filter having coefficients (a1, b1, a1) represented by the above formula (2) is subtracted from “1”, The filter has a zero at the spatial frequency fm. Considering the process of performing the filtering by this filter twice, it still has a zero at the spatial frequency fm, but the phase (sign) is not inverted. Such a filter is a symmetric 5-point FIR filtering filter, and a filter having a peak at the target spatial frequency fm can be configured by subtracting the filter from “1”.

  FIG. 5 shows the shapes of symmetric 3-point FIR filtering (hereinafter also referred to as “3-point FIR filtering”) and symmetric 5-point FIR filtering (hereinafter also referred to as “5-point FIR filtering”) as described above. This shows an example of the spatial frequency response characteristics.

  In most cases, the result of filtering by the FIR filter shown in FIG. 5 is mainly the result of extracting grid stripe components. This is because, as is apparent from FIG. 5, most of the low-frequency components out of effective image components mainly composed of low-frequency components are removed.

  However, as described above, it is also true that a significant amount of effective image components are included in the components extracted by FIR filtering. Originally, we would like to perform filtering with a filter having a steep selection characteristic (response characteristic) centered on the spatial frequency fm, but even if this is done, the frequency which constitutes the sudden fluctuation part included in the target image There is no change in the ingredients being included.

  Therefore, in order to solve the above problem, in the present embodiment, in the third step described above, a local envelope of the grid fringe component extracted by FIR filtering is obtained, and other than the grid fringe component is determined from the variation. By detecting a portion including a component that may cause an artifact, only the grid stripe component is stably extracted (created).

  The envelope of a general signal must be based on the Hilbert transform, but the envelope of a single sine wave has a response amplitude of “1” at that frequency, and the sine wave is 90 ° (π / 2). Is obtained by applying a spatial filter that causes phase fluctuations of the signal and taking the vector amplitude (square root of the sum of squares) of the result and the original signal.

When filtering with discrete data using a FIR filter whose phase varies by 90 °, the coefficients of the FIR filter are point-symmetric (odd functions). For example, when this coefficient is (−a3, 0, a3), in order to set the response to “1” at the spatial frequency fm, the coefficient (−a3, 0, a3) is
2 * a3 * sin (2πfmT) = 1
Must satisfy the formula

なる式（３）で表される解が得られる。

The solution represented by the following formula (3) is obtained.

  A vector amplitude between the signal sequence obtained by the FIR filter having the solution coefficient (−a3, 0, a3) represented by the above equation (3) and the original signal sequence is obtained.

  For example, FIG. 6A shows the result of filtering the image signal shown in FIG. 25A with the solution coefficients (a1, b1, a1) shown in the above equation (2). It is shown. As apparent from FIG. 6A, most of the grid stripe components are extracted.

  FIG. 7A shows the result of filtering the image signal shown in FIG. 26A with the solution coefficients (a1, b1, a1) shown in the above equation (2). It is shown.

  The waveforms shown in FIG. 6 (b) and FIG. 7 (b) (the waveforms shown by bold lines) are the above formulas (3) for the original image signals shown in FIG. 6 (a) and FIG. The envelope obtained by taking the square root of the square sum of the result of filtering having the solution coefficient (−a3, 0, a3) and the original image signal is shown.

  In particular, in the envelope shown in FIG. 7B, when attention is paid to the depression shown in FIG. 7C, an unsteady component is clearly present in the depression. This is because the extracted grid stripe component is abnormally extracted by simple filtering (including the edge component of the target image), and if the grid stripe component is subtracted from the target image signal, the processed target image signal This means that artifacts will occur.

  Therefore, in the present embodiment, a range indicating an abnormal numerical value is identified from the envelope obtained as described above, and the grid stripe component in the range is corrected (replaced) with an estimated value from the surrounding data string. ) That is, the grid stripe component is formed (created) using the characteristic of having a regular periodic component over the entire range, which is a characteristic of the original grid stripe component.

  The estimated value (predicted value) used in the correction is obtained from the statistical properties of the data around the range showing abnormal numerical values. For example, since the spatial frequency fm of the grid stripe component is known, this spatial frequency fm can be used as a statistical property.

  For example, with the spatial frequency fm of the grid stripe and the phase φ,

なる式（４）により表される正弦波を用いて、非定常部のグリッド縞成分を形成する。

The grid stripe component of the unsteady part is formed using the sine wave expressed by the following equation (4).

  For example, as the simplest method, there is a method in which two coefficients A and φ at a specific frequency are obtained from peripheral pixels using Fourier transform (Fourier series expansion).

  However, normal Fourier transform cannot be used due to problems such as the presence of defects (unsteady portions) in the data. Therefore, here, Fourier transform is generalized, and amplitude and phase information is obtained in the sense of least squares. Therefore, the above equation (4) is

なる式（５）に変形する。

The following equation (5) is transformed.

  In this case, when the data at the sampling point xi is “yi” (number of data points n) ({xi, yi; i = 0 to n−1}), the square error ε is

なる式（６）で表される。

It is represented by the following formula (6).

  At this time, it should be noted that as the component “xi, yi” used here, only data determined to be a stationary part from the above-described verification of the envelope component is selected. Then, parameters R and I that minimize the square error ε are obtained as follows.

  First,

なる式（７）は、

Equation (7)

なる式（７´）で書き表される。

This is expressed by the following equation (7 ′).

By solving the simultaneous equations of the above equation (7 ′), the parameters R and I can be obtained, and the phase φ and the amplitude A can be estimated simultaneously.
Here, if the data series divides the section of k / (2 · fm) (k is a positive integer) equally into m, the above equation (7 ′) is

なる式（８）で示されるように、特定の周波数の係数を求める離散フーリエ変換（フーリエ級数展開）となる。

As shown by the following equation (8), a discrete Fourier transform (Fourier series expansion) for obtaining a coefficient of a specific frequency is obtained.

  The unsteady state removed as inappropriate by obtaining the values of the parameters R and I by using the appropriate data around the unsteady portion according to the above formula (7 ′) or the above formula (8). Repair (replace) the part.

  As another repair method, there is a method in which a linear prediction model is considered and repair is performed by predicting sequentially using a linear prediction algorithm without specifying the spatial frequency of grid stripes.

  The signal waveform obtained by the above-described repair is generally a steady sine wave, and is a component that very well represents the grid stripe component.

  However, the above signal waveform (grid fringe component signal waveform) was originally the result of narrow span FIR filtering with the coefficients (a1, b1, a1) shown in equation (2) above, and is shown in FIG. It has such a response characteristic of the filter as described above and contains a large amount of image components other than the grid stripe components.

  Therefore, in the present embodiment, filtering for extracting only components near the spatial frequency fm of the grid stripes is performed on the signal waveform. This filtering is performed on a component in which an unsteady component has already been repaired by the above-described operation, and artifacts such as ringing due to the filtering do not occur.

  When the envelope generation described above is performed on the signal of the grid stripe component after the filtering, when a portion where a very small value (a value close to “0”) is observed is present, Originally, the grid stripe component is a portion that has not been observed for some reason (for example, X-rays are completely blocked or the sensor is saturated), and the grid stripe component does not originally exist. Therefore, information on this part is recorded and replaced with “0” after subsequent filtering. The result of this processing is subtracted from the target image signal as a grid stripe component.

  In the present embodiment, grid stripe component extraction processing is executed while sequentially extracting target line data from the target image signal one line at a time. When one line data is extracted, the average of several line data before and after the line data is extracted. It is also possible to extract the grid stripe component after obtaining the image, that is, after weakening the image component or enhancing the grid stripe component.

  As described above, the direction of the grid stripe and the direction in which the pixels of the sensor are arranged are generally parallel, and each component of the grid stripe of a certain line and the grid stripe of a line in the vicinity thereof is very Because it is very similar.

  Therefore, as a modification of the present embodiment, in order to reduce the number of calculation processes for extracting grid stripe components, the processing line is thinned out by using the above-mentioned very similar feature. The extracted grid stripe component can also be used as a grid stripe component of a line in the vicinity thereof. In other words, the grid fringe component extraction process is not performed on the line in the vicinity of the line from which the grid fringe component is extracted, and the subtraction process from the line data in the vicinity is performed using the grid fringe component extracted in the certain line. Can also be performed.

  In order to determine whether or not the above-described thinning is possible, in the first processing step described above, when measuring the spatial frequency of the grid stripes, the lines before and after being sampled or between the sampled lines are measured. The phase difference of the grid stripe may be checked to confirm that the grid stripe direction is not inclined with respect to the sensor pixel arrangement direction.

  As another form of the present embodiment, as described above, when the grid stripe component is removed from the target image signal, the irradiation field in the target image signal is recognized by detecting the signal intensity or the like. The above-described grid stripe component removal process may be performed only on the image data inside the irradiation field.

  By the way, for example, when acquiring an X-ray image with a solid-state image sensor, there is a problem of defective pixels as a problem peculiar to the solid-state image sensor formed by arranging a plurality of pixels. Due to the redundancy of image information (spatial low-frequency components are the main components), if there are only a small number of defective pixels, it is almost always a defect due to interpolation processing that replaces the average value of surrounding pixels. The pixel can be repaired (corrected).

  However, in general, prediction is necessary due to the statistical properties around defective pixels. For example, when the grid stripe is 50% or more of the Nyquist frequency as in the present embodiment, the prediction is reversed in the average interpolation.

  FIG. 8 shows a response function as filtering when an arbitrary point is a defective pixel in one dimension and interpolation is performed with the average of two pixels on both sides of the defective pixel. In FIG. 8, the spatial frequency is represented on the horizontal axis. As shown in FIG. 8, when the spatial frequency is low and it is 50% or less of the Nyquist frequency, the response is “positive”, that is, the phase is not reversed. On the other hand, when it becomes 50% or more of the Nyquist frequency, the phase is inverted, and an expected interpolation result cannot be obtained.

FIG. 9 shows an example of defective pixel interpolation.
In FIG. 9, a black dot represents a pixel value obtained from a normal pixel, and a point indicated by an arrow (“defective pixel position”) represents a defective pixel for which data is not obtained. Further, FIG. 9 shows a state in which each pixel data (data indicated by a black dot) vibrates finely due to the reflection of grid stripes.

  Also, the white circle point “A: average interpolation value” in FIG. 9 is a pixel value obtained by the conventional average interpolation, and “B: ideal interpolation value” in FIG. 9 is indicated. A certain white circle point is an interpolation value in consideration of grid stripes.

  In the present embodiment, the following two methods are performed in order to obtain an ideal interpolation value indicated by “B: ideal interpolation value” in FIG.

(Method 1: Linear prediction method)
In FIG. 9, the data of the defective pixel position is obtained by linear prediction from the surrounding pixels.

(Method 2)
By the above-described grid stripe removal process, the grid stripe component is removed from the original image signal, so that the main component (grid stripe component) of 50% or more of the Nyquist frequency is eliminated, and the interpolation by the average which is the conventional method is performed. Perform processing.

The outline of the linear prediction method (method 1) will be described below.
First, as image data (pixel data) to be processed, a data series {Xn, Xn-1, Xn-2,..., Xn-p} is given, and data Xn in “n”

なる差分方程式（９）で表されるものとする。

It is assumed that the difference equation (9)

  In the above equation (9), “εn” represents a white noise sequence, and “ai {i = 1,..., P}” represents a linear prediction coefficient. Such a sequence is called “autoregressive process (AR process) Xn”.

When the above equation (9) is rewritten using the delay operator Z ⁻¹ ,

なる式（１０）となる。

Equation (10) is obtained.

  However, the above formula (10) is

Therefore, the AR process Xn can be defined (spectrum estimation) as an output for an input εn of a linear filter having a pulse transfer function 1 / A (z ⁻¹ ).

  In addition, the above equation (9) is obtained by calculating the linear prediction coefficient ai {i = 1,..., P} from the reliable (n−1) -th pixel data. This indicates that the pixel data can be predicted.

  The prediction of the linear prediction coefficient ai {i = 1,..., P} can be performed by using maximum likelihood estimation (least square estimation) while assuming that the apparatus or system is stationary. That is, it is possible to obtain one that minimizes the power (dispersion) of εn. Since εn is an error obtained by least square estimation, it does not have a correlation component corresponding to the predicted order or less. Therefore, the prediction error εn obtained by prediction with the necessary and sufficient order p becomes white noise as defined in Equation 9.

  Since the variance of the prediction error εn is a square average (average value “0”), the function E [*] representing the average is

なる式（１１）で表される。

It is represented by the following formula (11).

In the above formula (11),
R (τ) = E [XmXm + τ] (covariance function: covariance)
In order to obtain the minimum value, if both sides are differentiated by a coefficient ak and set to “0”,

The following simultaneous equations (12) are obtained. This is what is called a normal equation or Yule-Walker equation.

  Actually, the autocorrelation R (*) is not calculated at all pixel points, but an estimated value calculated with a limited (given) number of pixels is used.

  For example, the Levinson algorithm which is a high-speed algorithm is used, but the Burg algorithm may be used. The Burg algorithm is an algorithm based on the maximum entropy method that can be obtained without directly calculating covariance (autocorrelation) with data having a smaller number of pixels. These algorithms mathematically match if the prediction error is a normal distribution and the number of pixels is large, but if the number of pixels is small, the Burg algorithm (an algorithm based on the maximum entropy method) is advantageous.

  The pixel data of the defective pixel expected from the pixels before and after the defective pixel is obtained using the coefficient ak obtained by the above calculation.

[Third Embodiment]
The present invention is applied to, for example, an X-ray image acquisition apparatus 100 as shown in FIG.

<Overall Configuration and Operation of X-ray Image Acquisition Apparatus 100>
The X-ray image acquisition apparatus 100 according to the present embodiment is an apparatus for acquiring an X-ray image for medical use (for image diagnosis, etc.). As shown in FIG. An X-ray generation unit 101 generated for a human body, a grid 103 for removing scattered X-rays from the subject 102, and a planar X-ray sensor (flat) for detecting the distribution of the X-ray dose transmitted through the subject 102 Panel sensor) 104, a controller 105 (CONT) of the X-ray generation unit 101, an analog / digital (A / D) converter 106 that converts an electrical signal output from the X-ray sensor 104 into digital data, an A / A memory 107 that temporarily stores digital data output from the D converter 106 as target image data, and a digital data acquired by radiography without emitting X-rays. A memory 108 that stores the data, a computing unit 109 that performs arithmetic processing using the data in the memory 108 on the target image data in the memory 107, and a conversion that converts the target image data after the arithmetic processing in the computing unit 109 A table (reference table: LookUpTable, also referred to as “LUT” hereinafter) 110, a memory 111 for storing gain pattern data for correcting variations in gain (sensitivity) for each pixel constituting the X-ray sensor 104, and the LUT 110. A computing unit 112 that performs arithmetic processing using the gain pattern data in the memory 111 on the output target image data after conversion; a memory 113 that temporarily stores the target image data after the computation result in the computing unit 112; Memory 11 that stores information (defective pixel position information, etc.) regarding defective pixels unique to the X-ray sensor 104 And a correction processing unit 115 that performs correction processing on the target image data stored in the memory 113 using information in the memory 114, and detects information on grid stripes in the target image data after correction processing in the memory 113. Grid fringe detecting unit 116, grid fringe component extracting unit 117 for extracting the grid fringe component from the corrected target image data in the memory 113 based on the information obtained by the grid fringe detecting unit 116, and the grid fringe component A memory 118 that temporarily stores the grid stripe component extracted by the extraction unit 117, a calculator 119 that subtracts the grid stripe component in the memory 118 from the target image data after correction processing in the memory 113, and a calculator 119 The memory 120 that temporarily stores the calculation result (target image data after removal of the grid stripe component) and the target image data in the memory 120 And an image processing unit 121 that performs image processing on the data and outputs the processed data.

  In the X-ray image acquisition apparatus 100 as described above, the controller 105 of the X-ray generation unit 101 starts X-ray emission at the X-ray generation unit 101 when a generation trigger is applied by an operator from an operation unit (not shown). Let

  Thereby, the X-ray generation unit 101 emits X-rays to the subject 102 that is a human body.

  X-rays radiated from the X-ray generation unit 101 pass through the subject 102 and reach the X-ray sensor 104 via the grid 103 that removes scattered X-rays from the subject 102.

  The X-ray sensor 104 has a configuration in which a plurality of detectors (pixels) for detecting the X-ray intensity are arranged in a matrix on a surface (image receiving surface) for detecting the distribution of the X-ray dose transmitted through the subject 102. An electric signal corresponding to the X-ray intensity obtained by the plurality of detectors (pixels) arranged in a matrix is output.

As the X-ray sensor 104, for example, the following sensors (1) and (2) are applicable.
Sensor (1):
A sensor that converts X-ray intensity into fluorescence once, and detects the fluorescence by photoelectric conversion with a plurality of detectors arranged in a matrix.
Sensor (2):
X-rays radiated to a specific object are photoelectrically converted in the object and free electrons are attracted by a uniform electric field to form a charge distribution, and the charge distribution is divided into a plurality of matrix-arranged charges. A sensor adapted to be converted into an electrical signal by a charge detector (capacitor).

  The A / D converter 106 digitizes and outputs the electrical signal output from the X-ray sensor 104.

  Specifically, the A / D converter 106 sequentially converts the electric signal output from the X-ray sensor 104 into digital data in synchronization with the X-ray emission of the X-ray generation unit 101 or the driving of the X-ray sensor 104. Convert to and output.

  In FIG. 10, one A / D converter 106 is provided. However, for example, a plurality of A / D converters may be provided and operated in parallel. As a result, the digital conversion speed can be increased, and the processing can proceed efficiently.

  The digital data output from the A / D converter 106 is temporarily stored in the memory 107 as target image data.

  Accordingly, the memory 107 stores digital image data (target image data) that is a set of a plurality of pixel data corresponding to a plurality of pixels constituting the X-ray sensor 104.

  The memory 108 stores in advance digital data acquired by imaging without emitting X-rays. This digital data is data for removing fixed pattern noise existing in an offset manner unique to the X-ray sensor 104 from the target image data stored in the memory 107. Therefore, in the X-ray image acquisition apparatus 100, imaging is performed in a state where the X-ray generation unit 101 does not emit X-rays, and the digital data acquired thereby is stored in the memory 108 as image data.

  The computing unit 109 obtains image data (X) stored in the memory 108 from each of a plurality of pixel data constituting target image data (image data obtained by X-rays transmitted through the subject 102) stored in the memory 107. A process of subtracting pixel data at corresponding positions among a plurality of pixel data constituting fixed pattern noise image data obtained by imaging without a line) is executed.

  The LUT 110 converts the target image data processed by the computing unit 109 into a value proportional to the logarithm and outputs it.

  The memory 111 stores gain pattern data for correcting variation in gain (sensitivity) of each pixel constituting the X-ray sensor 104 with respect to target image data converted by the LUT 110. For this reason, in the X-ray image acquisition apparatus 100, X-ray imaging is performed in the absence of the subject 102, and fixed pattern noise is removed from the obtained image data using digital data stored in the memory 108. Further, the data obtained by converting into a value proportional to the logarithmic value by the LUT 110 is stored in the memory 111 as gain pattern data.

  The computing unit 112 subtracts the gain pattern data in the memory 111 from the target image data output from the LUT 110 (corresponding to division if not logarithmically converted) and outputs the result.

  The target image data subjected to the subtraction processing by the calculator 112 is temporarily stored in the memory 113.

  Note that, when image data used for gain pattern data stored in the memory 111 is acquired, if image capturing is performed with the grid 103 attached, grid stripes will be reflected in the gain pattern data obtained as a result. Become. What is expected is that when the gain pattern data is subtracted from the target image data by the computing unit 112, the grid stripes reflected in the subject 102 are close to gain fluctuations, so that the grid stripe components are removed. It is possible that However, the gain pattern data obtained by shooting without the subject 102 is unlikely to be acquired every time an image is acquired (every actual shooting), and in most cases, is acquired once a day or less frequently. In addition, since the positional relationship between the X-ray generation unit 101 and the X-ray sensor 104 may change at each imaging, the grid stripe component is not removed by the above subtraction. Even if the positional relationship is unchanged, the imaging with the subject 102 and the imaging without the subject 102 generally differ in the X-ray scattered dose and the radiation quality, so the grid stripe contrast differs and the grid stripe component Is not removed by subtraction. In any case, if the direction of the grid 103 (the rotation angle of the foil constituting the grid with respect to the sensor pixel arrangement direction) does not change, the spatial frequency of the grid stripes does not change. Therefore, preferably, when gain pattern data is acquired, the grid 103 itself should be removed so that grid stripes are not included in the gain pattern data.

  The memory 114 stores information regarding defective pixels unique to the X-ray sensor 104 (defective pixel position information and the like).

  Specifically, for example, a flat X-ray sensor is generally manufactured by a semiconductor manufacturing technique, but the yield is not 100%, and a plurality of detectors (pixels) are not included due to some cause in the manufacturing process. Some are defective pixels that do not function properly as detectors, ie their output has no meaning. Here, the X-ray sensor 104 is inspected in advance in the manufacturing process or by means (not shown), and the position information of the defective pixel obtained as a result is stored in the memory 114.

  The correction processing unit 115 corrects the defective pixel data in the plurality of pixel data constituting the target image data stored in the memory 113 based on the position information of the defective pixel stored in the memory 114, and the corrected pixel The data is stored again in the corresponding position in the memory 113.

  The grid stripe detection unit 116 analyzes grid stripes on the target image data in the memory 113 (image data after correction processing by the correction processing unit 115), and the grid stripe spatial frequency fm and the grid stripe angle. Detect and output θ.

  The grid stripe component extraction unit 117 reads target image data (image data after correction processing by the correction processing unit 115) in the memory 113, and the grid stripe space obtained by the grid stripe detection unit 116 from the read image data. Based on the frequency fm and the grid stripe angle θ, a grid stripe component is extracted.

  The grid stripe component obtained by the grid stripe component extraction unit 117 is temporarily stored in the memory 118.

  The calculator 119 subtracts the grid stripe component stored in the memory 118 from the target image data in the memory 113 (image data after correction processing by the correction processing unit 115).

  The target image data from which the grid stripe component has been subtracted by the calculator 119 is temporarily stored in the memory 120.

  The image processing unit 121 performs image processing on the target image data in the memory 120 so that the observer can easily observe it.

Examples of the image processing here include the following processing.
-Random noise removal processing from the target image.
When the target image is displayed, the gradation is converted or a detailed part (for example, a predetermined spatial frequency component) is emphasized so that the density value is easy for the observer to see.
A portion unnecessary for the observer is cut out from the target image to reduce the information amount of the target image, or the target image information is compressed based on the redundancy.

  The target image data processed by the image processing unit 121 is displayed on a display unit, stored in a storage unit or a storage medium, and stored in a recording medium by an unillustrated means in the outside or the X-ray image acquisition apparatus 100. Processing such as recording or various analysis is performed.

<Specific Configuration and Operation of X-ray Image Acquisition Apparatus 100>
Here, in the above-described X-ray image acquisition apparatus 100, the following components that are considered to require specific description will be specifically described.
(1) Image data of grid stripe components stored in the memory 118.
(2) Correction processing of defective pixels by the correction processing unit 115.
(3) Grid stripe component detection and extraction processing by the grid stripe detection unit 116 and the grid stripe component extraction unit 117.

(1) Grid stripe component image data stored in the memory 118 The grid stripe component image data stored in the memory 118 is data subtracted from the target image data on which the grid stripe component is superimposed. If the configuration is such that the data stored in the memory 118 is separately stored in association with the target image data after subtraction, as in the above form, the original grid stripes are obtained from the target image data from which the grid stripes have been removed. The superimposed target image data can be reproduced. Thereby, for example, even if the target image data is damaged due to some trouble in the grid removal process, it is possible to restore the original target image data by the above reproduction process.

(2) Correction Processing of Defective Pixels by Correction Processing Unit 115 The correction processing unit 115 executes processing as described below, for example, by software using a microprocessor.

  11 to 14 show examples of pixel defect distributions in the X-ray sensor 104. FIG.

  Here, it is assumed that the pixel defect basically exists only in the width of one pixel. This is because an X-ray sensor having a plurality of adjacent pixel defects in a large lump is difficult to repair defective pixels and is not generally used.

  11 to 14, each square represents a pixel, and the black square represents a defective pixel. Further, in the lower part of the figure, the direction (vertical direction) of the grid stripes of the grid 103 is illustrated.

  First, the defective pixel shown in FIG. 11 is a basic pixel defect. As shown in FIG. 11, eight adjacent pixel components a1 to a8 exist around the defective pixel (black squares). Yes.

  FIG. 15 schematically shows the signal distribution on the spatial frequency axis of the grid stripe component in the target image in which the defective pixel shown in FIG. 11 is present and the grid stripe component is present.

  15, the horizontal axis represents the horizontal spatial frequency axis u of the target image, the vertical axis represents the vertical spatial frequency axis v of the target image, and both the spatial frequency axis u and the spatial frequency axis v. A “sampling frequency” that is the reciprocal of the pixel pitch and a “Nyquist frequency” that is a half value thereof are shown on the axis.

  Since the grid stripe vibrates in the horizontal direction of the target image and is constant in the vertical direction, the components of the grid stripe exist on the spatial frequency axis u as shown in FIG. (See white circle in the figure).

  In a normal image, the principal component is distributed in the spatial frequency region below the half value of the Nyquist frequency, and if there is no grid stripe component, the defective pixel is the average value of pixels on both sides adjacent to it. Can be interpolated. This is because the interpolation indicates the response function (characteristic) of the filtering shown in FIG.

  Therefore, in the case of the correction of the defective pixel shown in FIG. 11, since there is no grid stripe component in the vertical direction, the average of the pixels in the vertical direction, that is, the average of the pixel component a2 and the pixel component a6, or any one of the pixels. Depending on the component, almost satisfactory correction is possible.

  The defective pixel shown in FIG. 12 is a pixel defect that is connected horizontally (see black squares in the figure). Also in the case of a defective pixel having such a shape, the vertical direction of each pixel is a direction parallel to the grid stripes as in the defective pixel shown in FIG. Satisfactory correction is possible.

  The defective pixel shown in FIG. 13 is a pixel defect that is vertically connected (see black squares in the figure). In the case of a defective pixel having such a shape, there is no pixel having a reliable value above and below the target pixel except for the defective pixel at the upper end or the defective pixel at the lower end of the connected defective pixel. If defect correction is performed on a defective pixel in such a state using a simple average of adjacent left and right pixels, a correction result with an unexpected value as described with reference to FIG. 9 is obtained.

Therefore, in the present embodiment, the coefficient ak (k = 1 to P) is used by using the simultaneous equations as shown in the above equation (12) with the normal pixel components connected to the right or left side of the target defective pixel. ) Is determined from the left and right of the target defective pixel.
At this time, for example, the number of pixels to be used is about 20, and the order k is about 5.

  Then, the target defective pixel value Xn is predicted by the equation (9) with the coefficient ak, and the average of all the defective pixel values Xn obtained as a result is obtained. As a result, “B: ideal interpolation value” as shown in FIG. 9 is obtained.

  In the present embodiment, the coefficient ak (k = 1 to P) is obtained using the above equation (12). However, the present invention is not limited to this. For example, an algorithm called a maximum entropy method is used. Etc. may be used.

  The defective pixel shown in FIG. 14 is a defective pixel obtained by superimposing the state of the defective pixel shown in FIG. 12 and the state of the defective pixel shown in FIG. Among the defective pixels in this state, the pixel in question is a pixel at the intersection of the vertically connected defective pixel and the horizontally connected defective pixel (the pixel at the intersection of the crosses), that is, the pixel components a4, a12, and a18. , A24 is a defective pixel.

  The correction of the defective pixels in the state as shown in FIG. 14 is performed by correcting the linear defective pixels continuous in the horizontal direction by averaging the upper and lower pixel components, and the correction of the linear defective pixels continuous in the vertical direction is performed. , Using the simultaneous equations as described above.

Specifically, the following three methods (1) to (3) can be mentioned, but almost the same result can be obtained by any method.
(1) Correction of a defective pixel surrounded by the pixel components a4, a12, a18, and a24 is performed using an average value of upper and lower defect correction values (corrected defective pixel values).
(2) Correction of the defective pixel surrounded by the pixel components a4, a12, a18, and a24 is performed by solving the simultaneous equations of the above equation (12) using the values of the left and right pixels that are defect correction values.
(3) Correction of the defective pixel surrounded by the pixel components a4, a12, a18, and a24 is performed using the average value of the result of (1) and the result of (2).

  By executing the processing as described above, defective pixel data in a plurality of pixel data constituting the target image data stored in the memory 113 is corrected.

(3) Grid Stripe Component Detection and Extraction Processing by Grid Stripe Detection Unit 116 and Grid Stripe Component Extraction Unit 117 The grid stripe detection unit 116 reads a part of the target image data stored in the memory 113, and the read data Thus, the spectrum of the grid stripe included in the target image data is examined, and the spatial frequency fm and the angle θ of the grid stripe are detected. Information on the spatial frequency fm and the angle θ (hereinafter also referred to as “angle η”) of the grid stripes is used in the grid stripe component extraction process in the grid stripe component extraction unit 117 in the subsequent stage.

  FIGS. 16A to 16D are diagrams for explaining detection processing of the spatial frequency fm and the angle θ of the grid stripes.

  FIG. 16A shows an image of the entire target image, and “L1” to “L6” represent line positions from the top of the target image. The grid stripe detection unit 116 measures the spatial frequency fm of the grid stripe based on the result of Fourier transform of the lines L1 to L6. At this time, when detecting the spectrum of the grid stripe, the grid stripe detection unit 116 may use the average of several lines before and after each line L1 to L6 (or the average of the spectrum) in order to increase the detection capability. Good.

  FIGS. 16B to 16D each show a Fourier transform result in the line L1.

  That is, FIG. 16B shows the amplitude spectrum (or power spectrum), FIG. 16C shows the value of the real part that is the coefficient of the cosine wave of the Fourier transform result, and FIG. , The value of the imaginary part, which is the coefficient of the sine wave.

  FIG. 17 is a flowchart showing the above processing of the grid stripe detection unit 116.

  First, the grid stripe detection unit 116 clears a variable cumulation for obtaining an average of spectra (step S201).

  In addition, the grid fringe detection unit 116 clears the counter (variable) n of the number of lines that is a target for obtaining the average of the spectrum (step S202).

  Further, the grid stripe detection unit 116 initializes a variable i for selecting a processing target line (selected line) from the lines L1 to L6 shown in FIG. 16A to “1” (step S203). Thereby, in the first process, the line L1 is selected and processed as the target line.

  Then, the grid stripe detection unit 116 determines whether or not the processing of the next step S205 to step S215 has been executed for all the lines L1 to L6 of the target image (step S204).

  As a result of this determination, the process proceeds to step S216 described later only when the process is completed. If the process has not been completed yet, the process from the next step S205 is executed.

  If the processing is not completed as a result of the determination in step S204, first, the grid stripe detection unit 116 selects the line Li indicated by the variable i from the lines L1 to L6 of the target image, and the data (line data Li). ) Is acquired (step S205).

  Next, the grid stripe detection unit 116 performs a Fourier transform process such as a fast Fourier transform on the line data Li acquired in step S205 (step S206).

  Next, the grid stripe detection unit 116 obtains a power spectrum (or amplitude spectrum) from the Fourier transform result (spatial frequency domain data) in step S206 (step S207).

  Next, the grid stripe detection unit 116 determines whether or not there is a significant spectrum (peak value) indicating the grid stripe in the power spectrum acquired in step S207 (step S208).

  Specifically, for example, first, the absolute spatial frequency of the lead foil constituting the grid that causes the generation of grid stripes is known at the stage where the grid 103 is installed, so that the frequency is set to “fg”. By using it, the discrimination processing in step S208 can be performed accurately.

  That is, when the sampling pitch of the X-ray sensor 104 is “Ts”, the approximate spatial frequency fm where the grid stripes are generated is

なる条件式（１３）により特定できる。

It can be specified by the following conditional expression (13).

  At this time, in the above formula (13), when the condition shown in (3) is satisfied, the spatial frequency fm ′ obtained in (2) is used and the condition shown in (3) is not satisfied. In this case, (1) is executed with “J ← J + 1”.

  The accurate spatial frequency fm of the grid stripe should be present in the vicinity of “fm ′” obtained by the above equation (13), and it is determined whether or not a peak value (significant spectrum indicating the grid stripe) exists. In this case, if only the vicinity is searched, even if there are different peak values due to the influence of the image component or noise component caused by the subject, it is a significant spectrum showing grid stripes without being affected by it. Peak value can be detected.

  In addition, the grid 103 is manufactured fairly accurately (so that the spatial frequency of the lead foil is a predetermined fg over the entire grid), but the grid 103 and the X-ray sensor 104 Since the X-ray beam from the X-ray generation unit 101 has a cone beam shape, the shadow of the grid 103 is enlarged and reaches the X-ray sensor 104. For this reason, the spatial frequency fm of the grid fringe component actually present in the image is different from “fm ′” obtained by the above equation (13) and cannot be easily predicted.

  Accordingly, the frequency indicating the peak value among the frequencies in the vicinity of “fm ′” obtained by the above equation (13) is obtained as the spatial frequency fm.

  However, in the above case, it is necessary to determine whether or not the value corresponding to the obtained peak value is a significant peak value that actually exists stably. This determination is normally made based on the noise level. The noise level may be measured in advance, or an average value of components other than the peak value in the high band of the spectrum may be substituted. For example, if the ratio between the sum (or average value) of the power spectra of neighboring spectrum values indicating the peak value and the noise level is about 10 or more, it is determined that the peak value is usually significant and stable.

  As a result of the determination in step S208 as described above, if there is no significant peak value, the grid fringe detection unit 116 counts up the variable i to process the next line (step S217), and the step again Returning to S204, the subsequent processing steps are repeatedly executed.

  On the other hand, if there is a significant peak value as a result of the determination in step S208, the grid fringe detection unit 116 sets Pi as the spatial frequency indicating the peak value (step S209), and adds this to the variable accumulation (step S209). Step Ss210).

  Further, the grid stripe detection unit 116 obtains the phase of the grid stripe, sets the line position indicated by the phase and the variable i for the variable θn and the variable Mn (step S211 to step S214), and increments the variable n. (Step S215).

  Thereafter, in order to process the next line, the grid stripe detection unit 116 counts up the variable i (step S217), returns to step S204 again, and repeatedly executes the subsequent processing steps.

  As described above, when the processing of steps S205 to S215 is completed for all the lines L1 to L6, the grid fringe detection unit 116 changes the value of the current variable cumulation to the value of the current variable n ( By dividing by the number of significant peak values), an average grid fringe spatial frequency fm is obtained (step S216).

  In addition, after the processing of FIG. 17 is performed, the grid stripe detection unit 116 uses the variable θn obtained in step S213 and the line position Mn obtained in step S214 to obtain an average grid stripe angle η (angle θ )

That is, the grid stripe detection unit 116 calculates the difference in the grid stripe position by the phase difference (θi−θi + 1) between the phase θi of the i-th line Li and the phase θi + 1 of the (i + 1) th line Li + 1. With a spatial frequency fm,
{(Θi−θi + 1) / 2π} / fm
The line position difference (distance) between the line Li and the line Li + 1 is
(Mi + 1) -Mi
Using these results, the angle η of the grid stripe is

なる式（１４）により求める。

It calculates | requires by Formula (14) which becomes.

  And the grid stripe detection part 116 performs the process which extracts a grid stripe for every several lines, when the angle (eta) obtained by said Formula (14) is below a predetermined angle (when it is not inclined abnormally), On the other hand, when the angle η exceeds a predetermined angle (when the angle η is abnormally inclined), a process of extracting grid stripes is executed for each line.

  In the process executed by the grid stripe detection unit 116 described above, it is assumed that the direction (vertical or horizontal direction) of the grid stripe is known. For example, when the direction of the grid stripe is unknown, for example, The same processing is executed in advance in both the vertical direction and the horizontal direction, and the direction in which a significant peak value is detected may be set as a direction substantially orthogonal to the grid stripes.

  When the grid stripe detection unit 116 executes the processing as described above, the spatial frequency fm and the angle θ (angle η) of the grid stripe are obtained.

  The grid stripe component extraction unit 117 uses the spatial frequency fm and the angle θ (angle η) of the grid stripe obtained by the grid stripe detection unit 116, and the target image data including the grid stripe component stored in the memory 113. From this, the grid stripe component is extracted, and the grid stripe component is stored in the memory 118.

  FIG. 18 is a flowchart showing grid stripe component extraction processing in the grid stripe component extraction unit 117.

  First, the grid stripe component extraction unit 117 is given the grid stripe angle θ obtained by the grid stripe detection unit 116 and the spatial frequency fm of the grid stripe as processing parameters.

  The grid fringe component extraction unit 117 performs the processing of step S300 to step S321 described below based on the processing parameters (grid fringe angle θ and grid fringe spatial frequency fm).

  Note that when the value of the spatial frequency fm of the grid stripe given to the grid stripe component extraction unit 117 is “0”, the grid stripe does not exist on the target image, that is, the grid 103 is not used. In this case, the grid stripe component extraction unit 117 does not execute the processing shown in FIG. For example, the memory 118 stores “0” data as the grid stripe component in this case. Alternatively, since the calculator 119 does not function, the target image data stored in the memory 113 is stored in the memory 120 as it is.

  When the grid stripe angle θ and the spatial frequency fm of the grid stripe are given from the grid stripe detection unit 116 to the grid stripe component extraction unit 117, first, the grid stripe component extraction unit 117 uses the above formula (2). Then, the coefficients a1 and a2 (or the coefficients (a2, b2, c2, b2, a2) of the symmetric 5-point filter) are obtained from the spatial frequency fm (step S300). The FIR filter corresponding to the coefficient obtained here is “FIR1”.

  Next, the grid fringe component extraction unit 117 calculates the coefficient a3 from the spatial frequency fm using the above equation (3) (step S301). The FIR filter corresponding to the coefficient obtained here is “FIR2”.

  Next, the grid fringe component extraction unit 117 generates a window function centered on the spatial frequency fm in order to perform FIR filtering in the region of the spatial frequency fm (step S302).

  As the window function here, for example, a Gaussian distribution shape function centered on the spatial frequency fm can be applied.

  Next, based on the grid stripe angle θ, the grid stripe component extraction unit 117 determines a line range in which grid stripe extraction processing is performed on the line data constituting the target image (steps S303 to S305). ).

  Specifically, for example, the grid stripe component extraction unit 117 sets the reference value of the angle θ to “0.1 degree”, and the grid stripe angle θ obtained by the grid stripe detection unit 116 is less than the reference value of 0.1 degree. Is also larger or smaller (step S303). As a result of the determination, when the angle θ is smaller than the reference value 0.1 degree, the grid stripe component extraction unit 117 sets the variable skip to “5” to execute the process for every five lines. Is determined (step S304). On the other hand, when the angle θ is equal to or greater than the reference value 0.1 degree, the grid stripe component extraction unit 117 sets the variable skip to “0” to execute the process for all the lines. Determine (step S305).

  In steps S303 to S305, for example, not only the setting for the variable skip, but also a finer setting may be made based on the angle θ.

  After the process of step S304 or step S305, the grid stripe component extraction unit 117 initializes the variable count by setting the value of the variable skip set in step S304 or step S305 for the variable count. Is performed (step S306).

  Then, the grid stripe component extraction unit 117 determines whether or not the current variable count value is equal to or larger than the variable skip value, that is, whether or not the grid stripe extraction process for the target line data should be executed. (Step S307).

  If it is determined that the process is to be executed, the process proceeds to step S310. If not, the process proceeds to step S308. However, when this step S307 is first executed, it is always determined that the process is executed, so the process proceeds to the next step S310.

  As a result of the determination in step S307, in the case of processing execution (skip ≦ count), first, the grid fringe component extraction unit 117 first processes one line data (target line data) to be processed from target image data stored in the memory 113. ) Is acquired (step S310).

  In step S310, the target line data may be obtained as it is from the target image data. For example, an average value (moving average value) of several lines before and after the target line data is used as an actual processing target line. You may make it acquire as data.

  Next, the grid fringe component extraction unit 117 performs filtering using the FIR1 whose coefficient is determined in step S300 on the target line data acquired in step S310, and stores it in the line buffer 1 (not shown) (step). S311).

  As a result of this processing, the line buffer 1 stores image component data including grid stripe components.

  Next, the grid stripe component extraction unit 117 performs filtering using the FIR2 whose coefficient has been determined in step S301 on the line data stored in the line buffer 1, and stores the filtered data in the line buffer 2 (not shown) ( Step S312). As a result of this processing, the line buffer 2 stores data for obtaining the envelope of the grid stripe component.

  Next, the grid stripe component extraction unit 117 obtains an envelope of the grid stripe component (step S313).

  That is, the grid stripe component extraction unit 117 obtains the amplitude (that is, the square root of the sum of squares) of the vectors having the data in the line buffer 1 and the data in the line buffer 2 as components, and the result is obtained as (Not shown). The calculation here can be applied even to a calculation that does not take the square root due to the monotonic increase of the square root, and the same effect can be obtained.

  Next, the grid fringe component extraction unit 117 examines data in the line buffer 3, that is, envelope data, and determines an upper limit value th1 and a lower limit value th2 for detecting the abnormal data ( Step S314).

  Various methods can be applied to determine the upper limit value th1 and the lower limit value th2. For example, an average value and a standard deviation value are obtained, and n times the standard deviation value from the average value (n is, for example, “ 3) (a value about 3 ″) or a method in which the value shifted by more than the upper limit value th1 and the lower limit value th2 is obtained, or a histogram of envelope data in the line buffer 3 is obtained, and the upper limit value th1 and the lower limit value th2 centered on the mode The method etc. of determining are mentioned.

  Next, in the envelope data in the line buffer 3, the grid fringe component extraction unit 117 converts the data of the value th1 or more or the value th2 or less to abnormal data (abnormal data based on the fact that the image data fluctuates sharply). The data of the grid stripe component in the line buffer 1 corresponding to the abnormal data is estimated from the data around the abnormal data and rewritten (step S315). At this time, data rewriting is performed so that the entire grid stripe component is in a stable state showing a periodic variation pattern.

  Note that the processing in step S315 has already been described in detail using the above equation (7 ′) and the like, and thus the details thereof are omitted here.

  Next, the grid fringe component extraction unit 117 performs a Fourier transform process on the data of the grid fringe component in the line buffer 1 that has become an overall stable state by the process of step S315, and thereby the space of the grid fringe component. Frequency domain data is obtained (step S316).

  Note that the transformation processing in step S316 is not limited to Fourier transformation, and other orthogonal transformations such as cosine transformation can be applied.

  Next, the grid fringe component extraction unit 117 performs filtering by the window function centered on the spatial frequency fm obtained in step S302 on the spatial frequency domain data acquired in step S316 (step S317). As a result, the data of the grid stripe component more selectively represents the grid stripe.

  Next, the grid stripe component extraction unit 117 performs the inverse conversion process of the conversion in step S316 on the grid stripe component data after the filtering process in step S317, and uses the result as the actual grid stripe component data. (Step S318).

  Next, the grid stripe component extraction unit 117 stores the data of the grid stripe component acquired in step S318 at the corresponding position in the memory 118 (step S319).

  Then, the grid stripe component extraction unit 117 sets “0” for the variable count (step S320), and whether or not the processing from step S307 has been performed for all the line data constituting the target image data in the memory 113. Is determined (step S321).

  As a result of this determination, the present process is terminated only when the process is completed. When the process has not been completed, the process returns to step S307 and the subsequent process steps are repeatedly executed.

  On the other hand, as a result of the determination in step S307 described above, when the processing is not executed (skip> count), the grid stripe component extraction unit 117 copies the grid stripe component data acquired in the previous stage to the corresponding position in the memory 118. (Step S308), the variable count indicating the line to be copied is incremented (Step S309), the process returns to Step S307 again, and the subsequent processing steps are repeated.

[Fourth Embodiment]
The present invention is applied to, for example, an X-ray image acquisition apparatus 400 as shown in FIG.

  The X-ray image acquisition apparatus 400 of the present embodiment is different from the X-ray image acquisition apparatus 100 shown in FIG. 10 in the following configuration.

  In the X-ray image acquisition apparatus 400 of FIG. 19, the same reference numerals are given to components that operate in the same manner as the X-ray image acquisition apparatus 100 of FIG. 10, and detailed description thereof is omitted.

  In the X-ray image acquisition apparatus 100 shown in FIG. 10 described above, the correction processing unit 115 is configured to perform correction processing of defective pixels using the data in the memory 114 on the target image data in the memory 113. . On the other hand, in the X-ray image acquisition apparatus 400 according to the present embodiment, as shown in FIG. 19, the correction processing unit 115 has the target image data in the memory 120, that is, the target image after removal of the grid stripe component. The correction processing of the defective pixel using the data in the memory 114 is performed on the data.

  Therefore, according to the X-ray image acquisition apparatus 400 of the present embodiment, pixel defect correction in consideration of grid stripes is not necessary, and therefore, an average value of pixel values that are not peripheral defects as is conventionally known is used. Thus, simple defective pixel correction such as correcting defective pixels can be applied.

[Fifth Embodiment]
The present invention is applied to, for example, an X-ray image acquisition apparatus 500 as shown in FIG.

  The X-ray image acquisition apparatus 500 of the present embodiment is different from the X-ray image acquisition apparatus 100 shown in FIG. 10 in the following configuration.

  In the X-ray image acquisition apparatus 500 of FIG. 20, the same reference numerals are given to the components that operate in the same manner as the X-ray image acquisition apparatus 100 of FIG. 10, and detailed description thereof will be omitted.

  As shown in FIG. 20, the X-ray image acquisition apparatus 500 of the present embodiment further includes a detector (switch) that detects the mounting of the grid 103 in addition to the configuration of the X-ray image acquisition apparatus 100 of FIG. ) 122 is provided.

  The detection unit 122 supplies the detection result (grid mounting signal) of the mounting of the grid 103 to the correction processing unit 115 and the grid stripe detection unit 116, respectively.

  When the grid 103 is mounted based on the grid mounting signal from the detection unit 122, the correction processing unit 115 executes a defective pixel correction process in consideration of grid stripes as described in the third embodiment. . If not, the defective pixel is corrected by the average value of the pixel values of the peripheral non-defective pixels.

  Similarly, when the grid 103 is mounted on the grid stripe detection unit 116 based on the grid mounting signal from the detection unit 122, the grid stripe detection process (analysis process) as described in the third embodiment. Execute. However, if the grid 103 is not mounted based on the grid mounting signal, the grid stripe detection unit 116 immediately determines that there is no grid stripe without executing the grid stripe detection process, and processing corresponding thereto. Execute.

  As described above, in the X-ray image acquisition apparatus 500 according to the present embodiment, the detection unit 122 is provided, and the grid stripes are detected based on the detection result. Therefore, the grid is not attached. The time required for the process of detecting grid stripes can be reduced.

[Sixth Embodiment]
The present invention is applied to, for example, an X-ray image acquisition apparatus 600 as shown in FIG.

  The X-ray image acquisition apparatus 600 of the present embodiment is different from the X-ray image acquisition apparatus 100 shown in FIG. 10 in the following configuration.

  In the X-ray image acquisition apparatus 600 of FIG. 21, the same reference numerals are given to the components that operate in the same manner as the X-ray image acquisition apparatus 100 of FIG. 10, and detailed description thereof will be omitted.

  As shown in FIG. 21, the X-ray image acquisition apparatus 600 of the present embodiment further stores image data corresponding to the X-ray irradiation field in the configuration of the X-ray image acquisition apparatus 100 of FIG. For this purpose, a memory 123 is provided.

  In the memory 123, image data obtained by cutting out only image data of an area corresponding to an X-ray irradiation field (an area irradiated with X-rays in an image receiving area of the sensor) from the target image data stored in the memory 113 ( (Irradiation field image data) is stored, and the grid fringe component is detected and extracted from the irradiation field image data in the memory 123.

  Specifically, in X-ray imaging, in general, in order to avoid exposure of the subject 102 (here, the human body) other than the target region to be imaged, an irradiation field is provided at the exit of the X-ray generation tube of the X-ray generation unit 101. A diaphragm is provided. Thereby, only the necessary part of the subject 102 can be irradiated with X-rays.

  When the above-described irradiation field stop function is used, not all image data obtained from the X-ray sensor 104 is valid (necessary) and image data obtained by X-ray imaging corresponds to the X-ray irradiation field by the irradiation field stop. Only partial image data is valid (required).

  Therefore, in the present embodiment, X-rays are calculated from target image data in the memory 113 based on the X-ray intensity (pixel value) distribution, aperture shape, or other information by computer means (not shown). An effective partial image region (irradiation field) corresponding to the irradiation field is found, and only the data (irradiation field image data) of the irradiation field portion is stored in the memory 123.

  As described above, according to the present embodiment, not only the irradiation field image data in the memory 123, that is, the target image data but the entire image data of the necessary portion with the reduced amount of information is targeted. be able to.

  In this embodiment, the irradiation field is cut out from the target image after the defective pixel correction. For example, after the irradiation field is cut out, the defective pixel correction is performed on the irradiation field. Also good.

[Seventh Embodiment]
In the third embodiment, in the X-ray image acquisition apparatus 100 in FIG. 10 described above, the grid stripe component extraction processing in the grid stripe component extraction unit 117 is performed according to the flowchart in FIG.

  In the present embodiment, the grid stripe component extraction process in the grid stripe component extraction unit 117 is, for example, a process according to the flowchart shown in FIG.

  Note that, in the grid stripe component extraction process of FIG. 22, steps that are the same as the grid stripe component extraction process of FIG. 18 are denoted by the same reference numerals, and detailed description thereof is omitted.

  First, before the description of the grid stripe component extraction processing in the present embodiment, FIG. 23A shows a part of the target image including the grid stripe component. In FIG. The portion shown indicates a portion where the grid stripe component does not exist due to the presence of a substance that blocks X-rays.

  FIG. 23B shows the result of extracting the grid fringe component from the target image shown in FIG. 23A by the filtering according to the third embodiment. As shown in FIG. 23B, in the grid stripe component, an artifact appears in a portion corresponding to the portion indicated by “*” in the target image. Therefore, as described in the third embodiment, the portion is It is extracted from the envelope and inferred from the data before and after it so that the whole becomes a stable grid stripe. This result is shown in FIG. 23 (c). With such a stable grid stripe component, no new artifact is generated by a transformation process such as Fourier transform.

  In the third embodiment, the grid stripe component as shown in FIG. 23C is subtracted from the original image (target image) as shown in FIG. 23A to remove the grid stripe component. However, in this configuration, it is conceivable that a new fringe component appears in a portion where no grid fringe component originally exists (portion indicated by “*”).

  Therefore, in the present embodiment, as shown in FIG. 23D, in the grid stripe component as shown in FIG. 23C, the portion where the grid stripe component originally does not exist (the portion indicated by “*”). Is set to “0”.

  For this reason, in the present embodiment, the grid stripe component extraction unit 117 executes the grid stripe component extraction process according to the flowchart of FIG.

  That is, the grid fringe component extraction unit 117 replaces the portion of the stable grid fringe component data obtained by this process after estimation by step S315 with “0” after performing the process of step S318. (Step S700), and then proceeds to the next step S319. Thereby, it is possible to surely prevent a new fringe component from being generated by the grid fringe removal process even if the grid fringe component does not exist originally.

  In the first to seventh embodiments, the hardware is used. However, the entire apparatus can be realized by controlling the software.

  Another object of the present invention is to supply a storage medium storing software program codes for realizing the functions of the system or apparatus of the first to seventh embodiments to the system or apparatus, and to perform the computer of the system or apparatus. Needless to say, this can also be achieved by (or CPU or MPU) reading and executing the program code stored in the storage medium.

  In this case, the program code itself read from the storage medium realizes the functions of the first to seventh embodiments, and the program code and the storage medium storing the program code constitute the present invention. It will be.

  As a storage medium for supplying the program code, ROM, flexible disk, hard disk, optical disk, magneto-optical disk, CD-ROM, CD-R, magnetic tape, nonvolatile memory card, and the like can be used.

  Further, by executing the program code read by the computer, not only the functions of the first to seventh embodiments are realized, but also an OS running on the computer based on the instruction of the program code. Needless to say, the present invention also includes a case where the functions of the first to seventh embodiments are realized by performing part or all of the actual processing.

  Further, after the program code read from the storage medium is written to a memory provided in a function expansion board inserted into the computer or a function expansion unit connected to the computer, the function expansion is performed based on the instruction of the program code. A case where the CPU or the like provided in the board or the function expansion unit performs part or all of the actual processing and the functions of the first to seventh embodiments are realized by the processing is also included in one embodiment of the present invention. Needless to say, it is included.

FIG. 24 shows an example of the configuration of the computer function 800 described above.
As shown in FIG. 24, the computer function 800 relates to a CPU 801, a ROM 802, a RAM 803, a keyboard controller (KBC) 805 for controlling the keyboard (KB) 809, and a CRT display (CRT) 810 as a display unit. A CRT controller (CRTC) 806 that performs control, a disk controller (DKC) 807 that controls the hard disk (HD) 811 and the flexible disk (FD) 812, and a network interface controller (NIC) 808 for connection to the network 840 Are configured to be communicably connected to each other via a system bus 804.

  The CPU 801 generally controls each component connected to the system bus 804 by executing software stored in the ROM 802 or the HD 811 or software supplied from the FD 812.

  That is, the CPU 801 reads out a processing program according to a predetermined processing sequence from the ROM 802, the HD 811 or the FD 812 and executes it, thereby performing control for realizing the functions of the first to seventh embodiments.

The RAM 803 functions as a main memory or work area for the CPU 801.
The KBC 805 performs control based on an instruction or input from the KB 809 or a pointing device (not shown).
The CRTC 806 performs control related to display on the CRT 810.

  The DKC 807 accesses a boot program, various applications, edit files, user files, a network management program, a processing program for realizing the functions of the first to seventh embodiments, and the like to access the HD 811 or the FD 812. Control.

  The NIC 808 controls bidirectional data exchange with other devices or systems on the network 840.

[Operation and Effect of Radiation Image Processing Apparatus of Embodiment]
According to the radiological image processing apparatus of the embodiment described above, by removing only image components caused by the grid from the image data obtained by radiography through the grid without passing through the subject, the image sensor ( Correction image data suitable for correcting subject image data can be obtained based on sensitivity variations (gain variations) of a plurality of pixels constituting a solid-state imaging device.

  In addition, since only the image components caused by the grid can be removed from the calibration image data obtained by radiography through the grid without passing through the subject, calibration shooting can be performed without removing the grid. Very convenient.

  In addition, by using correction image data for performing image correction based on sensitivity (gain) variations of a plurality of pixels constituting the image sensor acquired in advance, an image caused by a grid can be obtained from calibration image data. Only components can be extracted (removed) with higher accuracy.

  Variations in sensitivity or gain (sensitivity image or gain image) of a plurality of pixels constituting the image sensor have less relevance between pixels and have steep fluctuation components. Therefore, the spatial frequency component of sensitivity variation is distributed over a wide spatial frequency region, unlike an image component of a normal subject. For this reason, the image component resulting from the grid and the image component with sensitivity variation are superimposed in both the frequency domain and the space, and separation may be difficult in practice. Even in such a case, sensitivity variation image data that does not include image components caused by the grid (for example, obtained by radiography without passing through the subject and the grid) is acquired and recorded in advance at a manufacturing factory or the like. If calibration image data including image components due to the grid is acquired, the sensitivity variation (gain) correction of the calibration image data is performed with the previously acquired sensitivity variation image data. Since steep fluctuation components due to sensitivity variations in image data are removed or reduced, it is possible to improve the accuracy of extracting only image components due to grids.

  Further, the image data for correction from which the image component due to the grid is removed is created, the image data obtained by photographing through the subject and the grid is corrected using the image data for correction, and the corrected image is obtained. By configuring to remove the image components caused by the grid from the data, the image components caused by the grid from the radiographic image data obtained by performing radiography through the fixed grid without moving during imaging Can be appropriately removed.

It is a block diagram of 1st Embodiment. It is a block diagram of the grid image removal part in 1st Embodiment. It is a block diagram of 2nd Embodiment. It is a figure explaining the phase change of the grid image resulting from the position change of a X-ray source. In 3rd-7th Embodiment, it is a figure for demonstrating the spatial frequency response characteristic of the filter for extracting a grid stripe component from a target image. It is a figure for demonstrating an example of the grid stripe component extracted from the target image by the filter of FIG. It is a figure for demonstrating the other example of the grid stripe component extracted from the target image by the filter of FIG. It is a figure for demonstrating the spatial frequency response characteristic of the interpolation process which correct | amends the defective pixel of a target image. It is a figure for demonstrating defect pixel correction | amendment with respect to the target image in which a grid stripe component exists. In 3rd Embodiment, it is a block diagram of the X-ray-image acquisition apparatus to which this invention is applied. In a 3rd embodiment, it is a figure for explaining an example (example 1) of defective pixel distribution in an object picture. In 3rd Embodiment, it is a figure for demonstrating an example (example 2) of defective pixel distribution in a target image. In 3rd Embodiment, it is a figure for demonstrating an example (example 3) of the said defective pixel distribution in a target image. In 3rd Embodiment, it is a figure for demonstrating an example (example 4) of the said defective pixel distribution in a target image. In 3rd Embodiment, it is a figure for demonstrating the spatial frequency distribution of the grid stripe component in a target image. In 3rd Embodiment, it is a figure for demonstrating the detection (analysis) of the grid stripe component in a target image. In 3rd Embodiment, it is a flowchart for demonstrating the detection (analysis) process of the grid stripe component in a target image. In 3rd Embodiment, it is a flowchart for demonstrating the process which extracts a grid stripe component from a target image. In 4th Embodiment, it is a block diagram of the X-ray-image acquisition apparatus to which this invention is applied. In 5th Embodiment, it is a block diagram of the X-ray-image acquisition apparatus to which this invention is applied. In 6th Embodiment, it is a block diagram of the X-ray-image acquisition apparatus to which this invention is applied. In 7th Embodiment, it is a flowchart for demonstrating the process which extracts a grid stripe component from a target image. In 7th Embodiment, it is a figure for demonstrating the process which extracts a grid stripe component, exemplifying image data. It is a block diagram which shows an example of the structure which reads the said program from the computer-readable storage medium or memory | storage part which recorded the program for making a computer implement | achieve the function of 1st-7th Embodiment. It is a figure for demonstrating an example of the effect of the filtering with respect to the image on which the grid stripe component obtained by radiography was superimposed. It is a figure for demonstrating the other example of the effect of the filtering with respect to the image which the grid fringe component obtained by radiography was superimposed.

Explanation of symbols

DESCRIPTION OF SYMBOLS 15 Storage part 16 Storage part 17 Grid image removal part 18 Storage part 19 Subtraction part 171 Storage part 172 Subtraction part 173 Grid component extraction part 174 Storage part

Claims (10)

A radiographic image processing apparatus for processing radiographic image data obtained by radiography using a grid for removing scattered radiation,
Means obtained by analyzing a spatial frequency of grid stripes generated in the radiation image data due to the grid;
A first means for performing band limitation processing based on the spatial frequency obtained by the analysis on the radiation image data and calculating image data obtained as a processing result as a first grid stripe component;
Means for calculating a value indicating an amplitude of the first grid stripe component;
Based on the value indicating the calculated amplitude, the value of the first grid fringe component is replaced with a value calculated from the value of the first grid fringe component, and the image data obtained by the replacement is replaced with a second value. A second means for calculating as a grid stripe component;
Means for subtracting the second grid stripe component from the radiographic image data.   In the second means, a second band limiting process based on the spatial frequency obtained by the analysis is performed on the image data obtained by the replacement, and the image data obtained as a result of the processing is converted into the second data. The radiation image processing apparatus according to claim 1, wherein the radiation image processing apparatus is calculated as a grid stripe component.   The radiographic image processing apparatus according to claim 2, wherein the band limited by the second band limiting process is narrower than the band limited by the band limiting process in the first means.   In the second means, when the value indicating the amplitude of the first grid stripe component is equal to or less than a predetermined value, the value of the second grid stripe component corresponding to the first grid stripe component The radiation image processing apparatus according to claim 1, wherein 0 is set to 0.   The radiographic image processing apparatus according to claim 1, wherein the band limiting process of the first means is performed by performing an FIR filtering process.   The radiographic image processing apparatus according to claim 1, wherein the value indicating the amplitude is an envelope. An image processing system configured such that a plurality of devices are communicably connected to each other,
An image processing system, wherein at least one of the plurality of devices has the function of the radiation image processing apparatus according to any one of claims 1 to 6. A radiographic image processing apparatus for processing radiographic image data obtained by radiography using a grid for removing scattered radiation,
Analyzing and obtaining a spatial frequency of grid stripes generated in the radiation image data due to the grid;
A first step of performing band limitation processing based on the spatial frequency obtained by the analysis on the radiation image data, and calculating image data obtained as a processing result as a first grid stripe component;
Calculating a value indicating the amplitude of the first grid stripe component;
Based on the value indicating the calculated amplitude, the value of the first grid fringe component is replaced with a value calculated from the value of the first grid fringe component, and the image data obtained by the replacement is replaced with a second value. A second step of calculating as a grid stripe component;
And a step of performing a process of subtracting the second grid stripe component from the radiation image data.   A computer-readable storage medium storing a program for causing a computer to realize the function of the radiation image processing apparatus according to claim 1 or the function of the image processing system according to claim 6. The program for making a computer perform the process of the radiographic image processing method of Claim 8.

Images