JP2004045052A - Apparatus and method for inspecting image data - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To reduce both the number of measurements on one inspection item and the amount of data to be captured in an inspecting apparatus in the inspection apparatus for processing a large amount of image data. <P>SOLUTION: From image data 1 inputted from a device 2d to be measured, local flaws are removed at a real-time rank filter 2h, and noise components are removed at an image data processing part 2j. Then data compression is performed within the range not affecting inspection, and compressed image data is stored in an image memory 2m. A subtractor 2k creates image data 3 representing the local flaws from the image data 1 and image data 2 from which the local flaws are removed, and the image data 3 is stored with its address in an image memory 2n in the case that it is determined by a comparator 2o that the image data 3 satisfies conditions from threshold value data. A test control device 2q inspects image characteristics through the use of image data 4 stored in the image memory 2m, and inspects flaws through the use of the image data 3 stored in the image memory 2n. <P>COPYRIGHT: (C)2004,JPO

Description

[0001]
TECHNICAL FIELD OF THE INVENTION
The present invention relates to an image data inspection apparatus and an image data inspection apparatus for performing an inspection in a manufacturing process of an image device such as a CCD / CMOS image sensor and a liquid crystal display panel, an inspection using captured image data, and the like to judge the quality of an image device to be inspected. About the method.
[0002]
[Prior art]
Examples of the inspection in the image device manufacturing process and the inspection using the image data captured by the image sensor include, for example, a flaw inspection for detecting a pixel having a characteristic abnormality from a measured value for each pixel, and collectively analyzing the measured value for each pixel. An image characteristic inspection or the like is performed. In the flaw inspection, for example, the output level of each pixel is measured, the image data is analyzed, and a white flaw inspection that detects a pixel whose output level is higher than that of the peripheral pixel is performed. There is a black scratch inspection to detect. The image characteristic inspection includes, for example, a sensitivity inspection in which an output level is measured when a constant amount of light is supplied and image data is analyzed for inspection, and a uniformity evaluation in which a sensitivity variation in a display screen is inspected.
[0003]
In the conventional image data inspection method, after the measured values of all pixels are fetched into an image data inspection device, an inspection process is performed using the fetched image data.
[0004]
FIGS. 4A to 4C are schematic diagrams each showing a processing flow of a conventional image data inspection method.
[0005]
First, image data is taken into the image data inspection device. FIG. 4A shows image data 1 in this case. Rank filtering is performed on the captured image data 1. In this rank filter processing, image data of a region designated around a pixel to be processed is ranked (ranked) in the order of larger value (or smaller value), and the image data of the pixel to be processed is assigned a specific ranking value (for example, Intermediate value). As a result, as shown in FIG. 4B, image data 2 for uniformity inspection from which local flaws have been removed is created.
[0006]
Next, based on the difference between the image data 1 shown in FIG. 4A and the image data 2 for uniformity inspection shown in FIG. 4B, only a local flaw as shown in FIG. The represented image data 3 for flaw inspection is created.
[0007]
Then, by analyzing the image data 2 and the image data 3 thus created, a uniformity inspection, a flaw inspection, and the like are performed.
[0008]
As described above, in the conventional image data inspection method, when image data is captured by the image data inspection device, data of all pixels may be used for all inspections, and the data measurement accuracy of all pixels may be reduced. It is necessary to match the inspection item that requires the highest accuracy. Therefore, during the data measurement of all pixels, data measurement is performed a plurality of times in order to eliminate random measurement value errors caused by the influence of clock jitter, external noise, etc. An averaging process divided by the number of times, a filtering process of removing singular data due to noise from a plurality of measured values, and the like are performed.
[0009]
In order to reduce such random measurement value error and noise, for example, Japanese Patent Application Laid-Open No. S61-188671 discloses a method in which captured image data is added for each predetermined block, and the total value is stored in a memory. An image processing apparatus configured as described above is disclosed. However, in the image processing apparatus disclosed in Japanese Patent Application Laid-Open No. S61-188671, a flaw inspection cannot be performed because local flaws are eliminated by adding image data for each block. In addition, since the change in data across blocks is reduced, image characteristic inspection cannot be performed sufficiently.
[0010]
Japanese Patent Application Laid-Open No. Hei 2-200060 discloses an image signal reading processing device configured to read the same portion of a document a plurality of times and average data in order to remove quantization error noise. ing. However, in the image signal reading processing apparatus disclosed in Japanese Patent Application Laid-Open No. 2-200060, after storing image data of all pixels in a memory, and averaging the data, the image signal reading processing apparatus is similar to the conventional image data inspection method. At the time when the data of all the pixels is taken into the apparatus, it is necessary to adjust the data measurement accuracy to the inspection item requiring the highest accuracy.
[0011]
[Problems to be solved by the invention]
The conventional image inspection data method shown in FIGS. 4 (a) to 4 (c), the image processing apparatus disclosed in JP-A-61-186871, and the image processing apparatus disclosed in JP-A-2-200060. In the image signal reading processing device, the processing of image data corresponding to the inspection item, the noise removal processing, etc. are performed after the data of all pixels are once taken into the image data inspection device. At the time when the image data is captured, it is necessary to match the measurement accuracy of the data with the inspection of individual pixels for which the highest accuracy is required.
[0012]
Therefore, in order to improve the data measurement accuracy of all the pixels, it is necessary to repeat the data measurement of all the pixels a plurality of times. There are problems such as an increase in the cost of the inspection device and an increase in the measurement time. In particular, in a device such as an image sensor or a display panel, the number of pixels has been rapidly increased, and it has become important issues to reduce the cost and shorten the measurement time of an image data inspection device.
[0013]
The present invention has been made in order to solve such problems of the related art, and can shorten the inspection measurement time when inspecting a large amount of image data, and can reduce the cost of the inspection apparatus. It is an object of the present invention to provide an image data inspection apparatus and an image data inspection method that can achieve the above.
[0014]
[Means for Solving the Problems]
An image data inspection apparatus according to the present invention is an inspection apparatus that processes a large amount of image data. The image data inspection apparatus performs preprocessing on image data to be inspected in accordance with an inspection item. A memory unit for storing the processed image data and an inspection unit for inspecting the image data stored in the memory unit are provided, thereby achieving the above object.
[0015]
The preprocessing unit can perform a process of removing a noise component from the inspection target image data.
[0016]
The pre-processing unit can perform a process of removing abnormal data that occurs randomly from the inspection target image data.
[0017]
The preprocessing unit can perform a process of narrowing down image data to be inspected to data necessary for the inspection.
[0018]
The preprocessing unit can perform a process of compressing the inspection target image data within a range that does not affect the inspection.
[0019]
Preferably, a plurality of the pre-processing units are provided, and a plurality of pre-processings are performed simultaneously.
[0020]
Preferably, a plurality of the memory units are provided, and the image data subjected to different preprocessing for each inspection item in the preprocessing unit is stored in different memory units.
[0021]
The inspection unit can separately inspect the image data from which the noise component has been removed by the preprocessing unit and the image data narrowed down to data necessary for the inspection by the preprocessing unit.
[0022]
The image data inspection method according to the present invention comprises the steps of: after removing abnormal data generated at random from the inspection target image data, performing predetermined processing and storing the data in a first image memory; The image data after the abnormal scratch data generated at random are removed is compared, and the comparison result satisfies the predetermined condition and the inspection target image data and the address of the image data are stored in the second image memory. Storing, and inspecting the image data stored in the first image memory or the image data stored in the second image memory, thereby achieving the above object. You.
[0023]
Hereinafter, the operation of the present invention will be described.
[0024]
In image data inspection such as inspection of imaging devices and inspection of captured image data, for example, it is necessary to measure individual pixels in a flaw inspection, and to collectively view a plurality of pixels in an image characteristic inspection such as sensitivity and uniformity. There is. Therefore, the measurement accuracy and the required data amount required for each inspection item are different from each other.
[0025]
Therefore, in the present invention, before storing the image data to be inspected in the memory unit, a process of removing a noise component, a process of removing abnormal data which occurs randomly, a process of narrowing down to data necessary for the inspection, and an inspection Pre-processing according to each inspection item, such as a process of compressing data within a range that does not affect the data, is performed to optimize the number of times data is captured and reduce the amount of captured data.
[0026]
For example, the flaw inspection is a test for a specific point having a flaw (black flaw, white flaw, etc.). Since an area without flaws is not an inspection target, it is narrowed down to a flaw-specific pixel having a flaw (or its surroundings) and measured. What is necessary is just to store data with high precision in the memory unit of the image data inspection apparatus. Therefore, after significantly reducing the number of pixels and storing and averaging the measured data in the memory unit a plurality of times, data accuracy can be improved.
[0027]
In the image characteristic inspection, data accuracy of each pixel is unnecessary, and a measured value can be determined with reference to peripheral pixel data. Therefore, since it is not necessary to perform the averaging process using the measurement data a plurality of times, the inspection time can be significantly reduced.
[0028]
As described above, by performing the preprocessing according to each inspection item, the inspection time can be shortened by reducing the number of data acquisitions, and the price of the image data inspection apparatus can be reduced by reducing the memory capacity for storing image data. Can be achieved.
[0029]
BEST MODE FOR CARRYING OUT THE INVENTION
Hereinafter, an embodiment of the present invention will be described with reference to the drawings.
[0030]
(Embodiment 1)
FIG. 1 is a functional block diagram showing a configuration of an image data inspection device according to an embodiment of the present invention.
[0031]
This image data inspection device 100 has a drive signal generator 2a and a driver circuit 2b. The drive signal generator 2a generates various clock signals and synchronization signals required for driving and measuring data of the device under test 2d such as an image sensor and a liquid crystal display panel to be inspected, and the driver circuit 2b generates the clock signal and the synchronization signal. It is converted into a signal waveform for driving the device 2d. By applying the drive signal waveform to the device under test 2d, the device under test 2d operates. The signal generated by the drive signal generator 2a is also supplied to an address signal generator 2c. The address signal generator 2c generates an image address of the device under test 2d based on the signal from the drive signal generator 2a. Data is generated. When the device under measurement 2d is an image device such as an image sensor, a light irradiation device, a power supply device, and the like are required for inspection, but are not directly related to the present invention. are doing.
[0032]
An analog signal (image data) output from the device under test 2d is supplied to the CDS / AGC 2f, sampled and held, amplified, and converted into a digital value by the ADC 2g. The image data 1 converted into a digital value by the ADC 2g is supplied to the real-time rank filter 2h. This image data 1 is the same as the image data 1 shown in FIG. In a recent CMOS image sensor or the like in which an ADC is built in the device, the CDS / AGC 2f and the ADC 2g are unnecessary.
[0033]
In the real-time rank filter 2h, the sequentially input image data 1 is processed in real time to remove randomly generated abnormal data, and is supplied as image data 2 to the image data processing unit 2j. The image data 2 processed by the real-time rank filter 2h is the same as the image data 2 shown in FIG. The real-time rank filter 2h is a functional block that performs rank processing on image data in a specific area around a pixel to be processed, and then performs rank filter processing to replace the specified ranking data in real time, using a known pipeline processing method. It is feasible.
[0034]
In the image data processing unit 2j, noise is removed from the image data 2 from which random abnormal data has been removed by rank filter processing within a range that does not affect sensitivity / image uniformity evaluation by low-pass filter processing, and data is emphasized. At the same time, data compression is performed within a range that does not affect the inspection by pixel data thinning or addition processing. In data compression, data is compressed to the minimum number of pixels required for inspection, for example, to the amount of data that can be stored in the storage capacity of the first image memory 2m. The image data 4 processed by the image data processing unit 2j is stored in the first image memory 2m.
[0035]
On the other hand, the image data 1 output from the ADC 2g is also supplied to the delay memory 2i at the same time. In the delay memory 2i, the image data 1 is delayed by the same period as the delay time generated by the real-time rank filter 2h, and supplied to the subtracter 2k as image data 1 '. The delay time generated by the delay memory 2i is also supplied to an address signal corrector 2e, which corrects a shift in address data caused by rank filter processing.
[0036]
Further, the image data 2 output from the real-time rank filter 2h is also supplied to the subtractor 2k.
[0037]
In the subtracter 2k, a difference between the input image data 2 and the image data 1 'is calculated in real time, and image data 3 is generated and supplied to the comparator 2o and the second image memory 2n. The image data 3 generated by the subtracter 2k is the same as the image data 3 shown in FIG.
[0038]
The threshold generator 21 generates threshold data for determining whether or not individual image data should be subjected to a flaw inspection. The threshold generator 21 generates different threshold data for each image area according to the address signal supplied from the address signal corrector 2e based on data written from the test control device 2q, preset data, and the like. It is supplied to the comparator 2o.
[0039]
The comparator 2o compares the image data 3 with the threshold data generated by the threshold generator 21 in real time with respect to a comparison area (image area) corresponding to the address signal from the address signal corrector 2e, and It is determined whether or not the inspection should be performed. For example, of the image data 3, it is determined that data having a value larger (or smaller) than the threshold data generated by the threshold generator 21 should be subjected to the flaw inspection. Then, a control signal for controlling the second image memory 2n is output according to the determination result.
[0040]
The second image memory 2n stores the image data 3 supplied from the subtractor 2k and the address signal of the image data supplied from the address signal corrector 2e according to the control signal from the comparator 2o. As a result, in the second image memory 2n, of the image data 3, only the data determined to be the target of the flaw inspection by the comparator 2o is stored.
[0041]
By the above operation, the first image memory 2m stores the image data necessary for the image characteristic inspection such as the sensitivity and the uniformity of the image, and the second image memory 2n obtains the measurement data of each pixel. The image data of the pixels required for the flaw inspection is stored.
[0042]
In the test control device 2q, the image data 4 is read from the first image memory 2m via the test control interface 2p, and the image characteristics such as sensitivity and image uniformity are inspected using the read image data 4. Done. In the test control device 2q, the image data 3 is read from the second image memory 2n via the test control interface 2p, and the read image data 3 is used to perform a flaw inspection such as a white flaw or a black flaw. Done. The test control device 2q can determine the quality of the device under test 2d based on these inspection results.
[0043]
The flaw inspection can be performed in real time by the test controller 2q via the test control interface 2p when the image data 3 is stored in the second image memory 2n. However, erroneous flaw inspection data may be stored in the second image memory 2n due to a randomly generated noise component. Therefore, the threshold data generated by the threshold generator 21 is required for flaw detection. It is desirable that the noise component is set lower than a certain level by an error due to noise, and the noise component is removed by comparing measured data a plurality of times. Hereinafter, a method of removing a noise component from the flaw inspection data will be described.
[0044]
FIG. 2 is a flowchart illustrating a processing flow for removing a randomly generated noise component from the flaw inspection data.
[0045]
First, as shown in FIG. 2A, the first image data 1-1 is supplied from the device under measurement 2d (or via the CDS / AGC 2f and the ADC 2f), and the first image data 1-1 shown in FIG. The second flaw inspection data 2-1 is extracted. At this time, since the threshold data is set lower than the level required for flaw detection by an error due to noise, the flaw candidates 1 to 3 of the flaw inspection data 2-1 contain noise components. May have been
[0046]
Next, as shown in FIG. 2C, the second image data 1-2 is supplied from the device under measurement 2d (or via the CDS / AGC 2f and the ADC 2f), as shown in FIG. 2D. The second flaw inspection data 2-2 is extracted. At this time, the flaw candidates 1 to 4 of the flaw inspection data 2-2 may contain noise components.
[0047]
Next, as shown in FIG. 2E, a pixel containing a noise component is detected by comparing the first flaw inspection data 2-1 and the second flaw inspection data 2-2.
[0048]
As described above, the threshold data generated by the threshold generator 21 is set lower than the level required for flaw detection by an error due to noise, and the threshold data is randomly generated by comparing a plurality of measurement data. The noise pixel to be determined can be determined and removed from the flaw inspection data. Further, by averaging a plurality of flaw inspection data, the accuracy of flaw inspection data other than noise components can be improved.
[0049]
The data for the flaw inspection generated in this way has a small amount of data because it does not cover all pixels, and the number of data measurements is minimized by the test controller 2q via the test control interface 2p. Can also be controlled. The control content of the test control device 2q can be realized by a general personal computer or the like.
[0050]
On the other hand, image characteristic inspections such as sensitivity and uniformity can be inspected by a single data measurement. After image data 4 is stored in the first image memory 2m, test control is performed immediately via the test control interface 2p. The image data 4 can be analyzed by the device 2q. Further, when a plurality of data measurements are required in the flaw inspection, the image data 4 can be generated and analyzed by effectively utilizing the data measurement time.
[0051]
As described above, according to the image data inspection apparatus of the present embodiment, the image data 3 and 4 preprocessed for each inspection item are stored in the individual image memories 2n and 2m, respectively. It is possible to perform data analysis narrowed down to data necessary for inspection sequentially or simultaneously.
[0052]
(Embodiment 2)
FIG. 3 is a functional block diagram illustrating a configuration of the image data inspection device 200 according to the present embodiment.
[0053]
This image data inspection device 200 has a general configuration of the image data inspection device 100 shown in FIG.
[0054]
This image data inspection device 200 has a data synchronization signal generator 4a and a control signal generator 4c. The data synchronization signal generator 4a generates data synchronization signals such as a synchronization signal and a clock signal required for driving an image device and measuring data, and synchronizes with the data synchronization signal to generate a plurality of image data to be measured. Are supplied to the pre-processing units 4h to 4j, respectively. In the control signal generator 4c, a control signal such as address data of an image is generated in synchronization with the data synchronization signal generated by the data synchronization signal generator 4a, and the control signal is generated via the test control interface 4p. , Preprocessing blocks 4h to 4j, image memories 4k to 4m, and the like.
[0055]
The pre-processing unit includes n first pre-processing units 4h to n-th pre-processing units 4j, each of which performs pre-processing according to an inspection item. For example, the first preprocessing unit 4h performs preprocessing for generating image characteristic inspection data (image data 4) such as sensitivity and uniformity evaluation described in the first embodiment, and performs second preprocessing. The unit 4i performs preprocessing for generating the flaw inspection data (image data 3) described in the first embodiment. The processing by each of the pre-processing units 4h to 4j is performed in parallel and simultaneously, and the processed image data is stored in the first image memory 4k to the n-th image memory 4m, respectively.
[0056]
The image data stored in each of the image memories 4k to 4m is read by the test control device 2q via the test control interface 2p, and the inspection is performed sequentially or simultaneously. As a result, the data is narrowed down to data necessary for the inspection, and data analysis can be performed using image data satisfying the measurement accuracy required for each.
[0057]
Note that the image data inspection device of the present embodiment can be configured by hardware, but it is also possible to apply a virtual signal processing structure using a parallel processing function in a data processing device such as a personal computer. It is.
[0058]
【The invention's effect】
As described above, according to the present invention, by performing pre-processing corresponding to an inspection item on image data, an inspection process is performed in real time, optimization of the number of times of repeated measurement according to each inspection item, It is possible to reduce the storage capacity of the image memory and the inspection processing time.
[0059]
For example, in the case of the flaw inspection data, by performing pre-processing that narrows down the data of the flaw inspection target pixels, the inspection can be performed while determining whether or not data measurement is necessary again. Can be realized. Further, even in the case of repeatedly performing data measurement, the amount of data to be processed can be significantly reduced, so that the inspection time can be reduced.
[0060]
In addition, since repetitive measurement is unnecessary and data of all pixels are unnecessary for data for image characteristic inspection such as sensitivity and image uniformity evaluation, preprocessing such as rank filter processing, low-pass processing, and data compression processing is performed. By doing so, the memory capacity used for storing data can be reduced to the number of pixels of the image device or less, and the cost of the inspection apparatus can be reduced.
[0061]
Furthermore, the same image data is not limited to the data for the flaw inspection and the data for the image characteristic inspection, but is pre-processed according to a plurality of inspection items, and the data required for the inspection is stored in the memory unit. Thereby, the efficiency of the inspection using the image data can be improved, and the price of the image data inspection device can be reduced.
[Brief description of the drawings]
FIG. 1 is a functional block diagram illustrating a configuration of an image data inspection device according to a first embodiment.
FIGS. 2A to 2E are flowcharts for explaining a processing flow for removing a randomly generated noise component from the flaw inspection data in the image data inspection apparatus according to the first embodiment;
FIG. 3 is a functional block diagram illustrating a configuration of an image data inspection device according to a second embodiment.
FIGS. 4A to 4C are flowcharts for explaining a processing flow of a conventional image data inspection method.
[Explanation of symbols]
2a Drive signal generator 2b Driver circuit 2c Address signal generator 2d Device under test 2e Address signal corrector 2f CDS / AGC
2g ADC
2h Real-time rank filter 2i Delay memory 2j Image data processing unit 2k Subtractor 21 Threshold generators 2m, 2n, 4k to 4m Image memory 2o Comparators 2p, 4p Test control interfaces 2q, 4q Test control device 4a Data synchronization signal generator 4c Control signal generator 4d Image data to be measured 4h to 4j Preprocessor 100, 200 Image data inspection device

Claims (9)

An inspection apparatus for processing a large amount of image data,
A preprocessing unit that performs preprocessing on the inspection target image data according to an inspection item;
A memory unit for storing image data after pre-processing by the pre-processing unit,
An image data inspection apparatus comprising: an inspection unit that performs an inspection using the image data stored in the memory unit. The image data inspection device according to claim 1, wherein the preprocessing unit performs a process of removing a noise component from the inspection target image data. The image data inspection device according to claim 1, wherein the preprocessing unit performs a process of removing abnormal data that occurs randomly from the inspection target image data. The image data inspection device according to claim 1, wherein the preprocessing unit performs a process of narrowing down the inspection target image data to data necessary for inspection. The image data inspection device according to claim 1, wherein the preprocessing unit performs a process of compressing the inspection target image data within a range that does not affect the inspection. The image data inspection apparatus according to claim 1, wherein a plurality of the pre-processing units are provided, and the plurality of pre-processings are performed simultaneously. The image data inspection apparatus according to claim 1, wherein a plurality of the memory units are provided, and the image data subjected to different preprocessing for each inspection item in the preprocessing unit is stored in different memory units. 7. The inspection unit according to claim 6, wherein the inspection unit separately inspects the image data from which the noise component has been removed by the preprocessing unit and the image data narrowed down to data necessary for the inspection by the preprocessing unit. Item 8. An image data inspection device according to Item 7. After removing abnormal data that occurs at random with respect to the inspection target image data, performing predetermined processing and storing the data in the first image memory;
The inspection target image data is compared with the image data after the abnormal data that has occurred at random has been removed, and the comparison result satisfies a predetermined condition. Storing the image data in the second image memory;
Performing a test using the image data stored in the first image memory or the image data stored in the second image memory.

Images