JP4815597B2 - Image processing method, image processing apparatus, and image processing program - Google Patents

Description

  The present invention provides an image processing for measuring the displacement, deformation, and distortion of an object by an image correlation method that obtains a correlation value between two images before and after the change of the object and regards the maximum point as a destination. The present invention relates to a method, an image processing apparatus, and an image processing program.

Conventionally, an image correlation method has been used as a method of obtaining displacement / deformation of an object by image processing. The image correlation method uses two images before and after the change, calculates the correlation value while scanning the partial area (reference image) of the other image data on one image data, and obtains the maximum correlation value. It is assumed that a reference image exists at a certain point, and in measuring the displacement amount of a two-dimensional image, a two-dimensional displacement amount can be derived by scanning the reference image in a two-dimensional direction. Such image measurement is effective when there is a large number of measurement objects or measurement ranges or measurement points that cannot be measured by the contact method.
By the way, in general, in the measurement by the image correlation method, the analysis is performed on the assumption that the measurement object is translated (for example, Patent Document 1). For this reason, it is difficult to perform analysis when the measurement target is greatly deformed and accompanied by deformation such as rotation, shear deformation, and expansion / contraction.
JP-A-8-166208

Therefore, the inventors have proposed a method of rotating the reference image by geometric transformation in the image correlation method (Non-Patent Document 1). In this method, the reference image is rotated by a certain angle and repeatedly scanned each time. In the conventional method that considers only translational motion, the analysis becomes impossible when the target is rotated about 10 degrees. Even when rotated 30 degrees, analysis is possible.
In addition, there has been proposed an invention in which affine transformation is performed on a target image when performing processing by an image correlation method (Patent Document 2). Here, the affine transformation is coordinate transformation by a combination of parallel movement, rotation, left / right reversal, enlargement / reduction, and shear. However, in Patent Document 2, only translation, rotation, and enlargement / reduction (same magnification in the x and y directions) are considered. The present invention scans the image data of an object to obtain a correlation value with a pre-registered model image, and repeats scanning a predetermined number of times while changing the affine transformation parameters to obtain the maximum correlation value. derive. Furthermore, in order to avoid performing affine transformation many times, the object is binarized to obtain a main shaft angle, and parameters of the affine transformation are determined based on it.
Hirose, Tashiro, Sonoda, Nomura, Kamiya, “Study on Displacement Measurement by Image Processing Using Two-Dimensional Geometric Transformation,” Proc. Japanese Patent Laid-Open No. 7-220026

However, in the method of Non-Patent Document 1, only rotation is considered when converting the reference image. Therefore, when a continuum is targeted, movement and rotation without deformation and a minute part of the continuum can be regarded as not distorted. Although it was possible to analyze small deformations, it was impossible to analyze in consideration of the distortion of a minute part.
In addition, the invention of Patent Document 2 recognizes an object and measures a position using a model image registered in advance, and cannot be used for deformation of a continuum. Furthermore, the method of determining the affine transformation parameters from the main shaft angle cannot be used because one main shaft angle cannot be determined when measuring deformation of a continuum.

  The present invention has been made in view of the above circumstances, and provides an image processing method, an image processing apparatus, and an image processing program that are mainly intended for a continuum and that can measure displacement, deformation, and distortion by an image correlation method. The purpose is to do.

  According to the first aspect of the present invention, the image data acquisition means accepts the pre-change image data and the post-change image data of the object, and the parameter acquisition means coordinates the analysis points on the pre-change image data. Parameter acquisition process that accepts numerical data of the size of the micro area centered on the analysis point and the moving amount analysis range, rotation angle analysis range, and distortion analysis range of the analysis point, and the rotation processing means rotates the micro area by affine transformation A rotation process for generating a mask rotated by a predetermined angle in the angular analysis range, and a correlation between the first correlation value calculation means and the changed image data while scanning each mask in the movement amount analysis range on the changed image data First correlation value calculation process for calculating the value and position angle detection process for detecting the position and angle at which the position angle detection means has the maximum correlation value The expansion / contraction processing means generates a secondary mask obtained by expanding / contracting the mask showing the maximum correlation value by a predetermined expansion / contraction ratio in the distortion analysis range by affine transformation, and the second correlation value calculating means changes with each secondary mask. A second correlation value calculation process for calculating a correlation value with the post-image data, an expansion / contraction ratio detection process for detecting an expansion / contraction ratio at which the expansion / contraction ratio detection means becomes the maximum correlation value, and a position where the state determination means gives the maximum correlation value, And a state determination process for determining the amount of movement of the analysis point, the rotation angle, and the distortion from the angle and the expansion / contraction rate, respectively.

  The invention according to claim 2 is characterized in that the shape of the minute region is circular.

  Furthermore, in the invention of claim 3, when analyzing the displacement, deformation, and distortion at a plurality of points on the continuum, the analysis range determining means analyzes the movement amount, rotation angle, and distortion of the current analysis point to be analyzed. For at least one of them, the analysis range determination process for determining a relative analysis range based on the movement amount, rotation angle and distortion value of the previous analysis point already analyzed at a point in the vicinity of the current analysis point is provided. Features.

  Further, the invention of claim 4 is an image processing apparatus, which is an image processing device for capturing an image of an object before and after change, and outputting the image as pre-change image data and post-change image data. And an image processing means for executing the image processing method according to 3.

  Further, the invention of claim 5 is an image processing program, characterized by causing a computer to execute the image processing method of claim 1, 2 or 3.

  According to the first aspect of the present invention, the displacement, deformation and distortion of the object can be measured from only the pre-change image data and the post-change image data without contacting the object. In particular, by applying expansion / contraction by affine transformation to a minute region during image correlation, it is possible to analyze large deformations with distortion of minute parts, which was impossible with the conventional method. This analysis can be performed with higher accuracy. Furthermore, it is also possible to directly grasp the stress at each point from the strain. Specific applications of the present invention include the measurement of large deformation of metals, analysis of structures such as bridges and buildings, and landslide disaster prediction systems such as landslides.

  According to the second aspect of the present invention, the shape of the minute region is circular, so that even when the minute region is rotated by affine transformation, the number of calculation points included in the region does not change and is not rotated. Can be analyzed under the same conditions.

  According to the third aspect of the present invention, by defining the analysis range of a certain point on the continuum from the analysis results of the points in the vicinity thereof, the analysis range can be limited and the amount of calculation can be reduced.

  According to the invention of claim 4 of the present invention, the user can measure by the image processing method of the present invention by imaging the object by the imaging means and processing the image data by the image processing means. it can.

  According to the invention of claim 5 of the present invention, the user can carry out the image processing method of the present invention with a general-purpose personal computer or the like.

  The present invention measures the displacement, deformation, and distortion of an object by an image correlation method that obtains the correlation value between two images before and after the change of the object and regards the maximum point as the destination. is there. Therefore, first, an affine transformation of an image used in the present invention will be described using equations.

The coordinate transformation formula by affine transformation is: (x, y) for the coordinates before transformation, (x ′, y ′) for the coordinates after transformation, (q _x , q _y ) for the center of rotation, θ for the rotation angle, The expansion / contraction rate in the x direction and the y direction is expressed by the equation (1) as (s _x , s _y ).
As a method for performing such coordinate conversion, there are a method called forward conversion and a method called reverse conversion. The forward conversion represented by the equation (1) is a method of interpolating the numerical value of the light intensity of the coordinates (x, y) of the image before conversion into the coordinates (x ′, y ′) after conversion, The inverse transformation represented by the following equation (2) is a method of referring to the light intensity to be interpolated to the coordinates (x ′, y ′) of the post-conversion image from the coordinates (x, y) of the pre-conversion image. Since the light intensity to be interpolated can be calculated by the interpolation method in the method using the inverse transformation, the inverse transformation is used here. As the interpolation method, there are nearest neighbor interpolation method, linear interpolation method, cubic convolution interpolation method and the like. Especially, the cubic convolution interpolation method has high accuracy and is suitable.

When the inverse matrix of Expression (2) is calculated, Expression (3) is obtained.
Further, when formula (3) is expanded, formula (4) is obtained.

  Next, the image correlation method used in the present invention will be described with reference to the drawings and equations. FIG. 1 is a conceptual diagram of the image processing method of the present invention. Here, a cantilever is assumed as the object T.

  First, on the pre-change image data 1, the coordinates of the analysis point 3 and the shape and size of the minute region 4 centered on the analysis point 3 are determined. The correlation value is calculated while shifting the minute region 4 on the changed image data 2 pixel by pixel, and the analysis point 3 is considered to exist at the point where the maximum correlation value is obtained.

  As the size of the minute region 4 is smaller, the spatial resolution is higher, but the measurement accuracy is lowered and an error is likely to occur. Therefore, it should be determined by the practitioner in consideration of the balance, but here it is 15 × 15 pixels. When the displacement of the analysis point 3 can be predicted to some extent, the shape may be short in the direction of displacement and long in the direction of not displacement.

  Further, the range (movement amount analysis range 5) for shifting the minute region 4 is the entire region on the post-change image data 2 when the displacement of the object T cannot be predicted at all, but when it can be predicted to some extent, the expected range. Shift within. For example, in the case shown in FIG. 1, it is considered that the beam is surely displaced downward. Therefore, the movement amount analysis range 5 may be limited to only the range below the analysis point 3.

Here, the correlation value of a certain point is obtained by taking the difference between a total of 9 pixels, that is, the pixel of the point and 8 pixels in the vicinity of the image before and after the change, and using the difference value. In Expression (5), the difference value of the light intensity at the point (x, y) of the pre-change image is g (x, y), and the difference value of the light intensity at the point (x, y) of the post-change image is g ′ ( As x, y), a correlation value S (Δx, Δy) is obtained. The displacement amounts when the maximum correlation value is shown are Δx and Δy.

  Note that the method described here is merely an example, and any method can be used as long as it can determine the degree of similarity between image data.

Next, rotation of the minute region 4 will be described. When the displacement of the object T is only a parallel movement or a minute rotation, it can be dealt with only by the above method. However, when a large rotational displacement is involved, the above method cannot obtain a maximum correlation value at an appropriate point. Therefore, it is necessary to rotate the minute region 4 by affine transformation to obtain a correlation. Here, the following formula (6) in which the expansion / contraction term is omitted in the formula (4) is used.

  Each minute region 4 rotated by a predetermined angle according to equation (6) is referred to as a mask 8 here. Here, the rotation center is the center of the minute region 4, that is, the coordinates of the analysis point 3. The correlation value is calculated while shifting the mask 8 generated in this way on the changed image data 2 one by one, and Δx, Δy and θ at which the maximum correlation value is obtained are the movement amount of the analysis point 3 and It is regarded as a rotation angle.

  Here, the range in which the minute region 4 is rotated (rotation angle analysis range 6) is the entire range of 0 ° to 360 ° when the rotation of the object T cannot be predicted at all. Rotate within the range. Further, the mask 8 may be generated for each fine angle (for example, every 1 °) from the beginning, but the mask 8 for each rough angle (for example, every 5 °) may be generated to calculate the respective correlation values. If the mask 8 for each fine angle is generated in the vicinity of the angle where the maximum correlation value is obtained and the correlation value is calculated again, the total number of masks 8 generated decreases, and the analysis time can be shortened.

  In addition, when rotating the micro area | region 4 in this way, the shape of the micro area | region 4 becomes a problem. That is, when the shape of the minute region 4 is a square, as shown in FIG. 2A, the number of calculation points for correlating with the deformed image is reduced by rotating (25 points → 17 points in the figure). ), Can not keep the same conditions. Therefore, as shown in FIG. 2B, if the minute region 4 is circular, the number of calculation points does not decrease even if it is rotated.

  Next, the expansion and contraction of the minute region 4 will be described. In addition to the above method, the distortion at the analysis point 3 can be obtained by considering the expansion and contraction of the minute region 4.

Here, each of the micro regions 4 in which the movement amount and the rotation angle are determined by the above-described method is expanded and contracted by a predetermined expansion / contraction rate according to the equation (4), and is referred to as a secondary mask 9 here. For the secondary mask 9 generated in this way, the correlation values with the image data 2 after change are calculated one by one, and s _x and s _y that give the maximum correlation value are regarded as distortions in the x and y directions. It is.

  Here, the range (distortion analysis range 7) in which the minute region 4 is expanded or contracted may be determined by comparing the pre-change image data 1 and the post-change image data 2 or the limit value of the distortion of the material of the object T ( If the value of the strain immediately before breaking) is known, it may be determined from that value.

  From the above, the movement amount, rotation angle, and distortion of the analysis point 3 were obtained. If this series of analyzes is performed over the entire area of the object T, the state of displacement, deformation and distortion of the entire object can be grasped.

  In the above method, the analysis point 3 and the minute region 4 are defined on the pre-change image data 1 and shifted on the post-change image data 2. On the contrary, the post-change image data 2 The analysis point 3 and the minute region 4 may be defined on the top, and may be shifted on the pre-change image data 1. According to the former method, since the movement destination of each point of the target object before the change can be grasped, it is suitable for examining how to deform when a force is applied to a certain object. On the other hand, according to the latter method, since the movement source of each point of the object T after the change can be grasped, it is suitable for examining which part is displaced and how during the landslide. .

  In order to actually carry out the image processing method of the present invention, the following image processing apparatus and image processing program are used.

  The image processing program of the present invention includes an input device composed of a keyboard and a mouse, an output device composed of a display, a CPU that executes instructions of the program in order, the program, data necessary for execution of the program, calculation results, etc. Is executed by a standard computer including a storage device for storing the computer, and the computer functions as various means. In addition, the imaging unit 200 in the image processing apparatus of the present invention includes a general digital camera, and the image processing unit 100 includes a computer that executes the program. FIG. 3 is a block diagram showing the configuration of the first embodiment of the image processing apparatus of the present invention, and FIG. 4 shows the image processing executed by the first embodiment of the image processing apparatus of the present invention. It is a flowchart which shows the flow of a method.

  The first process is an image data acquisition process, in which the imaging unit 200 captures a pre-change image and a post-change image of an object, and the image data acquisition unit 101 acquires the pre-change image data and post-change image data. To do. The resolution of the image data is not particularly limited as long as both are the same, but of course, the higher the resolution, the higher the analysis accuracy. Further, at this stage, a portion (a portion corresponding to the background) that is clearly not an object on the image data may be omitted.

  Next is a parameter acquisition process, in which the parameter acquisition unit 102 receives input of parameters from the input unit 300 by the user. The input unit 300 includes a general input device such as a keyboard and a mouse. Here, the coordinates of the analysis point, the size of the minute area centered on the analysis point, the analysis range of the analysis point movement amount (for example, ± 80 pixels in the x direction and the y direction), and the rotation angle analysis range (for example, ± 30) °) and a distortion analysis range (for example, 0.9 to 1.1 in the x direction and the y direction), and an actual size corresponding to one pixel is set. Multiple coordinates of analysis points can be specified. Also, the shape of the minute region can be specified, and according to the shape, the vertical and horizontal lengths are specified for a rectangle, and the radius is specified for a circle.

  Next, a filtering process is performed, and the filtering unit 151 performs a filtering process on the pre-change image data and the post-change image data to sharpen the image. Although there are various types of filters, a Laplacian filter is preferable here. By performing such processing, the edge of the image is enhanced and the measurement accuracy is improved.

  Next, a rotation process is performed, and the rotation processing unit 103 generates a mask obtained by rotating a minute region around the analysis point by affine transformation. For example, if the rotation angle analysis range is ± 30 ° and the rotation unit is 1 °, 60 masks are generated.

  The next is a first correlation value calculation process, in which the first correlation value calculation means 104 calculates a correlation value with the transformed image while shifting each mask pixel by pixel in the x and y directions. The range to be shifted is the previously set movement amount analysis range, and correlation values are obtained for all combinations of Δx, Δy, θ within the set range.

  Next is a position angle detection process. The position angle detection means 105 detects the maximum correlation value from the correlation values obtained in the first correlation value calculation process, and analyzes Δx, Δy, and θ giving the maximum correlation value. It is regarded as the amount of movement of the point and the rotation angle.

  The next is an expansion / contraction process, and the expansion / contraction processing means 106 generates a secondary mask that is expanded / contracted by affine transformation for the mask having the maximum correlation value obtained in the position angle detection process. For example, if the expansion / contraction rate analysis range is 0.9 to 1.1 in the x and y directions and the expansion / contraction unit is 0.01, 440 masks are generated.

Next is a second correlation value calculation process, and the second correlation value calculation means 107 calculates a correlation value between each mask and the transformed image. All s _x within the defined _range, the correlation values for the combination of s _y are determined.

The following is a scaling factor detection process, the expansion ratio detection means 108 detects the maximum correlation value from among the correlation values obtained in the second correlation value calculating step, analyzing the s _x, s _y giving the maximum correlation value Considered as point distortion.

The following is a state determination process, the state determining unit 109, [Delta] x that gives the maximum correlation value at position angle detection process, s _x which gives the maximum correlation value in Δy and θ as well as the expansion ratio detection _process, the s _y, the analysis points Considering the amount of movement, rotation angle, and distortion, the state of the analysis point after change is determined.

  Through the above process, the movement amount, rotation angle, and distortion for a certain analysis point are obtained. When a plurality of analysis points are designated, the rotation process to the state determination process are repeated by the number of analysis points.

  After the movement amount, rotation angle, and distortion for all analysis points are obtained, the last is an output process, and the output unit 400 outputs and displays the analysis result. The output unit 400 includes an output device such as a general display. The output format is a list that displays numerical data of the movement amount, rotation angle, and distortion of each analysis point, a display that displays displacement as a vector, a display that displays equal displacement lines in shades of color and colors, etc. Anything is acceptable.

  FIG. 5 is a block diagram showing the configuration of the second embodiment of the image processing apparatus of the present invention, and FIG. 6 shows the image processing executed by the second embodiment of the image processing apparatus of the present invention. It is a flowchart which shows the flow of a method.

  Here, the concept of the second embodiment will be described. In setting the analysis range, in the first embodiment, calculation is performed in the same analysis range for all analysis points, or an analysis range is set for each analysis point, but when considering displacement / deformation of a continuum, The analysis range can be determined efficiently. In other words, the amount of movement, rotation angle, and distortion of a point on the continuum are considered to be very close to the values of points near that point. Therefore, when determining the analysis range of a certain point, it is possible to determine a relative analysis range based on the amount of movement, rotation angle, and distortion of a point already analyzed at a nearby point. According to this method, since the relative analysis range can be narrowed at each analysis point on the continuum, the analysis time can be shortened. Below, only a different part from 1st embodiment is demonstrated.

  In the first embodiment, the movement amount analysis range, the rotation angle analysis range, and the distortion analysis range are set in the parameter acquisition process, but these are determined based on the current analysis point to be analyzed (for example, “± 80 pixels from the current analysis point”). On the other hand, in the second embodiment, in the analysis range determination process, the analysis range determination unit 110 receives the input value of the user's relative analysis range received by the parameter acquisition unit 102 and the current analysis point determined by the state determination unit 109. The movement amount, rotation angle, and distortion data of the previous analysis point that has already been analyzed at points in the vicinity of is read and the analysis range is set from these values (for example, the input value of the relative analysis range of the user is ± If the movement amount of 5 pixels and the previous analysis point is +10 pixels, “+5 to +15 pixels from the current analysis point”). If this method is used, the amount of movement, rotation angle, and distortion of the current analysis point are considered to be very close to those of the previous analysis point, so each analysis range can be set narrow, resulting in less computation. .

  When using the image correlation method, it is necessary that a characteristic pattern or mark be attached to the surface of the object. If the surface of the object is originally a random pattern, it may be used, but if there is no pattern or an inappropriate pattern, random dots may be applied. At this time, the number, distribution, and size of the dots affect the measurement accuracy. However, when applying a black circular dot on a white background, it is better to apply a dot with a random diameter than a dot with the same diameter. . In addition, the measurement accuracy is better when the ratio of the white background portion is higher than the black dot portion. Furthermore, you may apply | coat the dot of various colors (for example, red, green, blue).

  In addition, the minimum unit of displacement measurement by analysis of the image correlation method is one pixel, and basically it cannot be measured with a sensitivity lower than that. However, it is possible to measure with a sensitivity of about 1/10 of one pixel by two-dimensional sub-pixel processing. In this method, an approximate curved surface based on the correlation value of the maximum correlation point and its neighboring eight points is created by the least square method, and the true maximum correlation position is obtained in units of one pixel or less to obtain a displacement of one pixel or less. It is. Thereby, an appropriate result can be obtained even when the displacement is very small.

  In the above embodiment, the computer functions as each means by a program. However, a part or all of each means may be configured by dedicated hardware. Moreover, the order of each process may differ from the said embodiment in the range in which an equivalent result is obtained, and two or more processes may be performed in parallel.

  In the following, an analysis example in which the deformation of an object is actually measured by the image processing method of the present invention will be shown. The object to be measured was a metal vibration damper, which was deformed by applying force with a hydraulic cylinder, and an enlarged image was used for analysis. FIG. 7 shows an image used for the analysis, where (a) is a pre-change image and (b) is a post-change image. The lower part of the anti-vibration damper in the image is fixed, and the upper part is displaced to the right.

  First, for comparison purposes, analysis is performed using a method that considers only the translational motion of the measurement object and does not consider rotation. FIG. 8 shows the result of the analysis. (A) shows the amount of movement of each point in shades of color (the darker the color, the larger the amount of movement), and (b) shows the movement of each point. The state is represented by a vector. In (a), there are portions where the color shading changes discontinuously in some places, and it can be seen that an accurate analysis result is not obtained. Also in (b), there are vectors in different directions from the surroundings.

  Next, analysis by the method of the present invention is performed. In the method of the present invention, the movement amount and the rotation angle of the measurement object are obtained by analysis in consideration of translation and rotation, and the distortion of the measurement object is obtained by analysis in consideration of expansion and contraction. First, FIG. 9 shows the analysis result of the movement amount, and (a) and (b) represent the same as in the case of FIG. When analysis is performed in consideration of rotation, it can be seen that in (a), there are almost no discontinuous portions of color shading, and an accurate analysis result is obtained. Also in (b), a vector having a direction different from the surroundings is not seen.

  Further, for analysis considering expansion and contraction, three analysis points were selected here as shown in FIG. The analysis point A is a point where the distortion is considered to be small, the analysis point B is a point where the distortion is considered to be slightly generated, and the analysis point C is a point where the distortion is considered to be large. In the analysis, the expansion / contraction rate analysis range was 0.960 to 1.040 in the x and y directions, and the expansion / contraction unit was 0.005. (A), (b), and (c) of FIG. 11 show the calculation results of correlation values at points A, B, and C, respectively. From these, the maximum correlation value is obtained, and the expansion / contraction rate giving it is the distortion at each point.

Table 1 shows the coordinates of each point and the analysis results of the movement amount, the rotation angle, and the distortion. As for the coordinates, the upper left of the image is the origin, the right direction is the x direction, and the lower direction is the y direction. From the table, it is understood that the analysis point A has almost no distortion, the analysis point B has 2% distortion in the y direction, and the analysis point C has 2% distortion in the x direction and y direction, respectively.

  As described above, according to the method of the present invention, the movement amount and the rotation angle of the measurement object can be obtained with high accuracy. Furthermore, it is possible to analyze distortion that was impossible with the conventional method, and to grasp the state of the object in more detail.

The conceptual diagram of the image correlation used in the image processing method of this invention. Explanatory drawing which shows the relationship between the shape of a micro area | region, and rotation. 1 is a block diagram showing a configuration of a first embodiment of an image processing apparatus of the present invention. 3 is a flowchart showing a flow of an image processing method executed by the first embodiment of the image processing apparatus of the present invention. The block diagram which shows the structure of 2nd embodiment of the image processing apparatus of this invention. 9 is a flowchart showing a flow of an image processing method executed by a second embodiment of the image processing apparatus of the present invention. Actual image used for analysis. The figure which shows the analysis result of the movement amount by the conventional method. The figure which shows the analysis result of the moving amount | distance by the method of this invention. Explanatory drawing which shows the position of an analysis point. The graph which shows the calculation result of the correlation value in the analysis of distortion.

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 Image data before change 2 Image data after change 3 Analysis point 4 Micro area 5 Movement amount analysis range 6 Rotation angle analysis range 7 Distortion analysis range 8 Mask 9 Secondary mask T Target object 100 Image processing means 101 Image data acquisition means 102 Parameter Acquisition means 103 Rotation processing means 104 First correlation value calculation means 105 Position angle detection means 106 Expansion / contraction processing means 107 Second correlation value calculation means 108 Expansion / contraction rate detection means 109 State determination means 110 Analysis range determination means 200 Imaging means

Claims (5)

An image data acquisition process in which the image data acquisition means (101) receives the pre-change image data (1) and the post-change image data (2) of the object (T);
The parameter acquisition means (102) coordinates the analysis point (3) on the pre-change image data (1), the size of the minute region (4) around the analysis point (3), and the amount of movement of the analysis point (3). A parameter acquisition process for receiving numerical data of an analysis range (5), a rotation angle analysis range (6), and a strain analysis range (7);
A rotation processing step in which the rotation processing means (103) generates a mask (8) by rotating the minute region (4) by a predetermined angle in the rotation angle analysis range (6) by affine transformation;
The first correlation value calculation means (104) calculates the correlation value with the after-change image data (2) while scanning each mask (8) in the movement amount analysis range (5) on the after-change image data (2). A first correlation value calculation process;
A position angle detection process in which the position angle detection means (105) detects the position and angle at which the maximum correlation value is obtained;
An expansion / contraction processing step (106) for generating a secondary mask (9) in which the expansion / contraction processing means (106) expands / contracts the mask (8) showing the maximum correlation value by a predetermined expansion / contraction ratio in the distortion analysis range (7) by affine transformation;
A second correlation value calculating unit (107) for calculating a correlation value between each secondary mask (9) and the post-change image data (2);
An expansion / contraction rate detection process in which the expansion / contraction rate detection means (108) detects an expansion / contraction rate that has a maximum correlation value;
A state determination process in which the state determination means (109) determines the amount of movement, rotation angle, and distortion of the analysis point from the position, angle, and expansion / contraction rate at which the maximum correlation value is given;
An image processing method comprising:   The image processing method according to claim 1, wherein the minute region has a circular shape. When analyzing displacement, deformation and distortion of multiple points on a continuum,
At least one of the movement amount, rotation angle, and distortion of the current analysis point to be analyzed by the analysis range determining means (110) is the movement amount and rotation of the previous analysis point that has already been analyzed at a point in the vicinity of the current analysis point. The image processing method according to claim 1, further comprising an analysis range determination step for determining a relative analysis range based on the angle and the distortion value. Imaging means (200) for capturing an image of the object (T) before and after the change and outputting it as image data before the change (1) and image data after the change (2);
Image processing means (100) for executing the image processing method according to claim 1, 2 or 3,
An image processing apparatus comprising: On the computer,
An image processing program for executing the image processing method according to claim 1, 2 or 3.