JP2017068349A - Image processing method and control program - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a technology of segmenting an image into a plurality of regions on the basis of feature values, which alleviates effects of defects on results of the region segmentation, without changing processing methods in accordance with types of defects.SOLUTION: An image processing method comprises a first step (S105-S106) in which feature values are derived for each of blocks into which an image has been segmented, a second step (S107) in which the feature values for each block are corrected on the basis of the feature values of a plurality of peripheral blocks which are in the periphery of the block, and a third step (S108) in which the image is region-segmented into a plurality of regions according to the corrected feature values of each block.SELECTED DRAWING: Figure 5

Description

  The present invention relates to an image processing method for dividing an image into a plurality of regions based on feature value.

  In the technical field of area division processing that obtains feature values from an image and divides the image into a plurality of areas according to the value, partial defects such as noise, dirt, and distortion mixed in the image affect the processing result. May have an effect. For this reason, it is necessary to remove such defects from the image. As a technique aiming at removing defects such as noise from an image, there is one described in Patent Document 1, for example. In this technique, the gradation level difference caused by highlighting is eliminated by performing correction processing that multiplies the luminance value of the target pixel by a coefficient corresponding to the difference between the average value of the luminance values of surrounding pixels. . As described above, it is generally performed to correct the defect by correcting the pixel value (luminance value) of the pixels constituting the image.

JP 2011-022656 A

  The prior art described above is specialized in eliminating gradation steps due to highlights. However, when correcting an image itself as described above, it is necessary to prepare a correction method according to the type of defect. This is because the influence on the feature value varies depending on the type of defect. For this reason, for example, in an image including a plurality of types of defects, it is necessary to perform individual correction processing for each type of defect, and the time required for the processing becomes longer, or it is necessary to calculate the feature amount by excessive correction. Problems such as loss of image information occur. In addition, it is difficult to predict in advance what kind of defect will appear in the image.

  The present invention has been made in view of the above problems, and in the technique of dividing an image into a plurality of regions based on feature value, the influence of defects on the result of region division without changing the processing method depending on the type of defect. The purpose is to provide a technology capable of suppressing the problem.

  In one aspect of the image processing method according to the present invention, in order to achieve the above object, a first step of obtaining a feature amount for each block obtained by dividing an image into a plurality of blocks, and a feature amount for each block A second step of correcting based on feature value values of a plurality of peripheral blocks in the vicinity of the image, and a third step of dividing the image into a plurality of regions according to the feature value values of the corrected blocks And.

  In the invention configured as described above, from the viewpoint of satisfactorily performing region division based on the value of the feature amount, the feature obtained from the image is not corrected by correcting the luminance value of the image. A correction is made to the quantity. Specifically, the feature value for each block obtained by dividing the image into a plurality of blocks is corrected based on the feature value of the blocks around the block.

  If the feature value calculated for each block includes a contribution due to a defect included in the image of the block, it is considered that such a contribution appears in the surrounding blocks including the same defect as well. Therefore, it is possible to reduce the influence of the defect by performing correction based on the feature value of the peripheral block with respect to the feature value of the target block to be corrected. Even when the influence on the value of the feature amount varies depending on the type of defect, the correction is performed using the feature amount values including the influence, so that it is not necessary to change the correction method according to the type of defect.

  Another aspect of the present invention is a control program that causes a computer to execute each step of the above-described image processing method. The series of processes described above can be executed using hardware resources included in a computer device having a general configuration, and provided as a control program executable by such a computer, The processing of the invention can be executed.

  As described above, in the present invention, since the feature value obtained for each block obtained by dividing the image into a plurality of blocks is corrected based on the feature values of the peripheral blocks, not the brightness value of the image, Regardless of the type, it is possible to reduce the influence of defects on the feature value. Therefore, it is possible to suppress the influence of defects included in the image on the result of area division.

It is a figure which shows schematic structure of the imaging device which can apply the image processing method concerning this invention. It is a figure which shows typically the example of the image imaged with this imaging device. It is a figure explaining the principle of the correction process in this embodiment. It is a figure for demonstrating the specific content of a correction process. It is a flowchart which shows the image processing content performed by this imaging device.

  FIG. 1 is a diagram showing a schematic configuration of an imaging apparatus to which an image processing method according to the present invention can be applied. This imaging device 1 is a cell, cell colony, bacteria, etc. (hereinafter referred to as “cell etc.”) cultured in a liquid injected into a recess called a well W formed on the upper surface of the well plate WP. This is an apparatus for imaging a raw sample.

  The well plate WP is generally used in the fields of drug discovery and biological science. The well plate WP is formed in a cylindrical shape with a substantially circular cross section on the top surface of a flat plate, and the bottom surface is transparent and flat. Are provided. Although the number of wells W in one well plate WP is arbitrary, for example, 96 (12 × 8 matrix arrangement) can be used. The diameter and depth of each well W is typically about several mm. The size of the well plate and the number of wells targeted by the imaging apparatus 1 are not limited to these, and are arbitrary. For example, those having 12 to 384 holes are generally used. The imaging apparatus 1 can be used not only for imaging a well plate having a plurality of wells but also for imaging cells or the like cultured in a flat container called a dish.

  A predetermined amount of liquid as a medium is injected into each well W of the well plate WP, and cells or the like cultured under predetermined culture conditions in the liquid are imaging targets of the imaging apparatus 1. The medium may be one to which an appropriate reagent has been added, or it may be in a liquid state and gelled after being placed in the well W. As will be described later, in the imaging apparatus 1, for example, cells cultured on the inner bottom surface of the well W can be targeted for imaging. Commonly used liquid volume is about 50 to 200 microliters.

  The imaging apparatus 1 includes a holder 11 that holds a well plate WP in a substantially horizontal posture by contacting a lower surface peripheral portion of a well plate WP that holds a sample in each well W together with a liquid, and an illumination unit disposed above the holder 11. 12, an imaging unit 13 disposed below the holder 11, and a control unit 14 having a CPU 141 that controls the operation of each unit.

  The illumination unit 12 emits appropriate diffused light (for example, white light) toward the well plate WP held by the holder 11. More specifically, for example, a combination of a white LED (Light Emitting Diode) light source as a light source and a diffusion plate can be used as the illumination unit 12. The illumination unit 12 illuminates cells and the like in the well W provided on the well plate WP from above.

  An imaging unit 13 is provided below the well plate WP held by the holder 11. In the imaging unit 13, an imaging optical system (not shown) is disposed immediately below the well plate WP, and the optical axis of the imaging optical system is directed in the vertical direction. FIG. 1 is a side view, and the vertical direction in the figure represents the vertical direction.

  The imaging unit 13 images a cell or the like in the well W. Specifically, light emitted from the illumination unit 12 and incident on the liquid from above the well W illuminates the object to be imaged, and light transmitted downward from the bottom surface of the well W is not illustrated via the imaging optical system. Is incident on the light receiving surface. An image of the imaging target imaged on the light receiving surface of the imaging device by the imaging optical system is captured by the imaging device. As the image sensor, for example, a CCD sensor or a CMOS sensor can be used, and either a two-dimensional image sensor or a one-dimensional image sensor may be used.

  The imaging unit 13 can be moved in the horizontal direction and the vertical direction by a mechanical control unit 146 provided in the control unit 14. Specifically, the mechanical control unit 146 moves the imaging unit 13 in the horizontal direction based on a control command from the CPU 141, so that the imaging unit 13 moves in the horizontal direction with respect to the well W. The focus is adjusted by moving in the vertical direction. When the imaging target in one well W is imaged, the mechanical control unit 146 positions the imaging unit 13 in the horizontal direction so that the optical axis coincides with the center of the well W. When the imaging device of the imaging unit 13 is a one-dimensional image sensor, a two-dimensional image can be captured by scanning the imaging unit 13 in a direction orthogonal to the longitudinal direction of the image sensor. With such an imaging method, it is possible to perform non-contact, non-destructive and non-invasive imaging on cells or the like that are imaging targets, and suppress damage to cells or the like due to imaging.

  In addition, when moving the imaging unit 13 in the horizontal direction, the mechanical control unit 146 moves the illumination unit 12 integrally with the imaging unit 13 as indicated by a dotted arrow in the drawing. That is, the illuminating unit 12 is arranged so that the optical center thereof substantially coincides with the optical axis of the imaging unit 13, and moves in conjunction with the imaging unit 13 when moving in the horizontal direction. As a result, regardless of which well W is imaged, the center of the well W and the optical center of the illumination unit 12 are always located on the optical axis of the imaging unit 13, and the illumination condition for each well W is made constant. The imaging conditions can be maintained satisfactorily.

  An image signal output from the image sensor of the imaging unit 13 is sent to the control unit 14. That is, the image signal is input to an AD converter (A / D) 143 provided in the control unit 14 and converted into digital image data. The CPU 141 executes image processing as appropriate based on the received image data. The control unit 14 further includes an image memory 144 for storing and storing image data, and a memory 145 for storing and storing programs to be executed by the CPU 141 and data generated by the CPU 141. It may be integral. The CPU 141 executes various control processes described later by executing a control program stored in the memory 145.

  In addition, the control unit 14 is provided with an interface (IF) unit 142. The interface unit 142 accepts operation input from the user and presents information such as processing results to the user, and exchanges data with an external device connected via a communication line. It has a function. In order to realize the user interface function, the interface unit 142 is connected to an input receiving unit 147 that receives an operation input from the user and a display unit 148 that displays and outputs a message to the user, a processing result, and the like.

  The control unit 14 may be a dedicated device equipped with the hardware described above, and a control program for realizing processing functions to be described later is incorporated into a general-purpose processing device such as a personal computer or a workstation. It may be. That is, a general-purpose computer device can be used as the control unit 14 of the imaging device 1. In the case of using a general-purpose processing device, it is sufficient that the imaging device 1 has a minimum control function for operating each unit such as the imaging unit 13.

  FIG. 2 is a diagram schematically illustrating an example of an image captured by the imaging apparatus. Specifically, FIG. 2A illustrates an example of an ideal image captured by the imaging apparatus 1, and FIG. 2B illustrates an example of an image including a defect. Hereinafter, the horizontal direction of the image captured by the imaging unit 13 is referred to as “X direction”, and the vertical direction is referred to as “Y direction”. As shown in FIG. 2A, the image Ia is imaged so that the entire well W fits in a rectangular area corresponding to the imaging field of the imaging unit 13. In the substantially circular region Rw corresponding to the inside of the well W in the image Ia, cells C and the like are scattered in the uniformly distributed medium M.

  As will be described later, in the imaging apparatus 1, the well region Rw is divided into a region of the medium M and a region of cells C, etc. by an appropriate region dividing process using the feature value calculated for each pixel constituting the image. Divided. However, as shown in FIG. 2B, the actually imaged image Ib may include some image defect D. Image defects can occur due to various causes such as noise, dirt, distortion, and the like.

  Such a defect may affect the calculation of the feature value and the result of the area division process based on the result. Therefore, correction processing is required to reduce the influence of defects appearing in the image on the region division processing. Hereinafter, correction processing in this embodiment will be described.

  FIG. 3 is a diagram for explaining the principle of correction processing in this embodiment. The upper part of FIG. 3 schematically shows a part of an image including the defect D like the image Ib shown in FIG. Consider a case where, for example, a feature amount F related to edge strength is obtained from this image Ic. The lower part of FIG. 3 schematically shows a change in the feature amount F obtained for each pixel located on the line segment AA in the image Ic. As shown in the figure, the value of the feature amount F is remarkably increased in the pixel corresponding to the outline of the cell C or the like, and a value different from the other portion appears in the portion corresponding to the defect D.

  Here, the value of the feature amount F is smaller in the portion of the defect D than in the other portions. However, depending on the type of the defect D and the type of the feature amount F, the value may be larger than the other portion or almost. There may be no difference.

  Since it is considered that the influence on the feature amount due to the defect D appears in a similar manner between the pixels corresponding to the defect D, the feature amount value of the target pixel is set to, for example, the surrounding values like the values F1 and F2 shown in the figure. A feature that is not affected by the presence or absence of the defect D is expressed as a relative value based on the feature value of the pixel (more specifically, an appropriate number of pixels selected in order from the closest distance to the target pixel). It can be an amount. This corresponds to correcting the feature value of the pixel of interest based on the feature value of other pixels around the pixel. Since the actual image has a two-dimensional spread, the specific contents of the correction processing are as follows.

  FIG. 4 is a diagram for explaining the specific contents of the correction processing. More specifically, FIG. 4A is a diagram showing the relationship between the pixel of interest and peripheral pixels used for correction, and FIG. 4B is a diagram showing the value of the feature amount F used for correction. It is. As shown in FIG. 4 (a), the pixel of interest P (x, y) whose X coordinate value is x and Y coordinate value is y is the center, and N pixels (N is 3 or more) in the X direction and Y direction, respectively. Considering a rectangular region Rr having a size of (integer), the feature amount of the pixel of interest P (x, y) is corrected using the feature value of the pixel in the region Rr. Hereinafter, each pixel in the region Rr excluding the target pixel P (x, y) is referred to as a “peripheral pixel” for the pixel P (x, y).

The pixel at the upper left corner of the rectangular area Rr is denoted by P (x1, y1), the pixel at the upper right corner is denoted by P (x2, y1), the pixel at the lower left corner is denoted by P (x1, y2), and the pixel at the lower right corner is denoted by P. It is represented by (x2, y2). here,
x1 = x- (N-1) / 2
x2 = x + (N-1) / 2
y1 = y- (N-1) / 2
y2 = y + (N-1) / 2
It is.

As shown in FIG. 4B, when the value of the feature amount before correction of the pixel P (x, y) is represented by the symbol Fr (x, y), the feature after correction of the pixel P (x, y). The quantity Fc (x, y) is:
Fc (x, y) = Fr (x, y) −Fa (x, y)
Defined by Here, Fa (x, y) is an average value of the feature values Fr (i, j) of the respective pixels P (i, j) in the rectangular region Rr, and the following formula (1) or formula (2) Represented by either

  Expression (1) represents an average value of feature amounts of peripheral pixels other than the target pixel P (x, y) among the pixels in the rectangular region Rr. Expression (2) represents the average value of the feature values of all the pixels in the rectangular region Rr including both the target pixel P (x, y) and the surrounding pixels. As the average value Fa (x, y) of the feature values of peripheral pixels serving as a reference when correcting the feature value Fr (x, y) of the target pixel P (x, y), the above formulas (1) and (2 ) May be used.

In the above, after correcting the difference between the feature value Fr (x, y) of the target pixel P (x, y) and the average value Fa (x, y) of the feature values of the surrounding pixels (and the target pixel). However, a value obtained by scaling by adding an appropriate offset value to the difference value or multiplying by a coefficient may be used as the corrected feature value. In general, the corrected feature value Fc (x, y) is expressed by the following equation using constants a, b, and c (where a ≠ 0):
Fc (x, y) = a · {Fr (x, y) −Fa (x, y) + b} + c
Can be represented by That is, the corrected feature quantity Fc (x, y) is generally defined as a linear function having a variable between the feature quantity Fr (x, y) before correction and the surrounding pixel (and the target pixel) as a variable. it can.

  For example, when it is necessary to visualize the feature value by mapping the corrected feature value on the XY coordinate plane, the corrected feature value is appropriately scaled as described above. It becomes possible to enhance the visualization effect.

  There are various possible feature quantities for which such correction functions effectively, but in the experiments of the present inventor, the third order with the pixel value (luminance value) as the third coordinate axis orthogonal to the X coordinate and Y coordinate is used. Good correction in the feature value called "NormalStDev", which is obtained as the standard deviation of the normal vector direction at each pixel position on the curved surface drawn by the pixel value profile obtained by plotting the pixel value of each pixel in the original coordinate space It has been confirmed that results can be obtained.

  More generally, when the feature value can be expressed by a continuous function having the pixel value of each pixel as a variable, the above correction method based on the average value of the feature values of the surrounding pixels functions effectively. In particular, when the feature amount is expressed as a convex function or a linear function with the pixel value as a variable, a good correction result can be obtained.

  FIG. 5 is a flowchart showing the contents of image processing executed by this imaging apparatus. This process is a process for distinguishing the area of the cells and the like from the image of the well W carrying the cells and the like, and the control program stored in the memory 145 is stored in the CPU 141 by the CPU 141. This is implemented by causing each part of the apparatus to perform a predetermined operation. The control program may be incorporated in the memory 145 in advance, or may be read from an appropriate storage medium and written to the memory 145 via the interface unit 142 as necessary.

  First, imaging of the well W is performed by the imaging unit 13 (step S101), and the well region Rw (FIG. 2) is cut out from the obtained original image (step S102). A well image representing the well region Rw is displayed on the display unit 148 (step S103). The content of the image displayed at this time is the same as that of the captured original image. When the original image includes a defect, the image is displayed in a state including the defect. The user can grasp the presence / absence of a defect and the extent thereof from the displayed image.

  Subsequently, an operation input from the user regarding the correction size information is received by the input receiving unit 147 (step S104). The above correction algorithm itself can be applied regardless of the type and degree of the defect, but the size of the rectangular region Rr that determines the range of the peripheral pixels used for the correction needs to be appropriately selected according to the appearance of the defect. Specifically, the size of the rectangular region Rr must be larger than the size of such a texture so that information on useful textures (for example, contour lines of cells or the like) in the image is not lost. On the other hand, in order to effectively remove the influence of the defect by subtracting the average value of the feature amount, it is preferable that the defect appears to appear uniformly in the region. The size of the rectangular region Rr must be smaller than the size of the region in which appears.

  In order to deal with various defects, the size of the rectangular region Rr is preferably variable, and more effective correction can be performed by accepting information for determining the size from the user as correction size information. Can be done. For example, the value of the number of pixels N that determines the length of one side of the rectangular region Rr can be received as the correction size information.

  Subsequently, the well image is divided into a plurality of small blocks (step S105), and the feature amount of each block is calculated (step S106). In the above example, the feature amount is obtained in units of pixels, which corresponds to the case where one pixel is one block. Instead, one block may include several pixels. For example, the image can be divided into blocks of a size corresponding to the size of the texture of interest in the image. When correction is performed using the feature amount obtained in units of blocks, it is desirable that the blocks have the same size.

  With respect to the feature amount of each block thus obtained, the value of the feature amount of the block is corrected based on the value of the feature amount of the surrounding blocks using the above-described method (step S107). By performing an appropriate region dividing process based on the corrected feature value of each block (step S108), the well image is divided into a region corresponding to cells C and the like and a region corresponding to the medium M. Since the influence of the image defect is reduced in the corrected feature amount, the influence of the image defect on the result of the area division process is suppressed. The region corresponding to the cell etc. C may be classified into a plurality of types according to the type and shape of the cell etc. C (for example, normal cells and undifferentiated cells etc.).

  As a method of area division, a known machine learning method using a feature value, for example, a random forest method can be used. Alternatively, for example, a method may be used in which a threshold value in several steps is set for the corrected feature value, and the region is divided according to the magnitude relationship with the threshold value.

  As described above, in the above embodiment, steps S105 to S106 in FIG. 5 correspond to the “first step” of the present invention, while steps S107 and S108 are the “second step” and “second step” of the present invention, respectively. This corresponds to “3 steps”. In the above embodiment, each of the pixels constituting the image corresponds to a “block” of the present invention. Of the pixels in the rectangular region Rr, the pixels other than the target pixel P (x, y) correspond to the “peripheral block” of the present invention.

  Further, among the average value Fa (x, y) of the feature quantity, what is represented by the expression (1) corresponds to the “first average value” of the present invention, while what is represented by the expression (2) is the present value. This corresponds to the “second average value” of the invention.

  The present invention is not limited to the above-described embodiment, and various modifications other than those described above can be made without departing from the spirit of the present invention. For example, the above embodiment is an imaging apparatus that can execute the image processing method according to the present invention, but the apparatus that executes the image processing method of the present invention may not have an imaging function. As described above, the image processing method according to the present invention can be executed by using hardware resources of a personal computer or workstation having a general configuration. The present invention may be realized by mounting a control program describing each processing step of the image processing method. In this case, an image captured by an external imaging device, an image previously stored in a database or a storage medium, and the like are processing targets of the present invention.

  Further, in the above-described embodiment, correction is performed by receiving a setting input of correction size information from the user, but the present invention is not limited to this. For example, the configuration may be such that the degree of defects is estimated by appropriate image processing, and the size of the rectangular region Rr is automatically determined according to the result. Further, for example, there may be a mode in which several sizes of the rectangular region Rr are prepared in advance and a correction result using them is presented to the user.

  In the above embodiment, each of the pixels constituting the image is a “block” of the present invention, which is the smallest in principle as a “block”. In this case, it is possible to perform correction while maintaining image information of a fine texture represented by a pixel unit included in the image. However, the block size may be larger depending on the application. For example, for the purpose of counting cells of substantially the same size scattered in the wells, a block composed of a plurality of pixels corresponding to the size of such cells may be set.

  Moreover, although the said embodiment performs an area | region division process with respect to the image which imaged the well which carries a cell etc., the image used as a process target in this invention uses such a raw sample as the imaging target object. It is not limited and various images can be used.

  As described above, the specific embodiment has been described by way of example. In the image processing method of the present invention, for example, in the second step, the value of the feature value for each block is set to the value of the feature value of each peripheral block of the block. You may make it correct | amend based on the 1st average value which is an average value, or the 2nd average value which is the average value of each feature-value of the said block and the surrounding block of the said block. According to such a configuration, it is possible to cancel the contribution due to defects that have the same degree of influence on the feature quantities of the block and the peripheral blocks.

  Further, for example, in the second step, the corrected feature value of the block may be a value corresponding to the difference between the feature value of the block and the first average value or the second average value. According to such a configuration, a value obtained by subtracting a value due to the influence of a defect from a value of the feature value of the block can be used as a corrected feature value.

  In this case, in the second step, a value represented by a linear function whose variable is the difference between the block feature value and the first average value or the second average value is used as the corrected block feature value. It may be a value. In this way, the corrected feature value can be appropriately scaled, and the feature value in which the influence of the defect is reduced can be output in a mode according to the purpose.

  For example, a plurality of blocks may have the same size. According to such a configuration, the feature value of each block has the same weight as each other, and it is not necessary to give a special weight in the calculation at the time of correction, so that the processing can be simplified. .

  Further, for example, in the second step, a block other than the block that is included in a rectangle of a predetermined size including a plurality of blocks and centered on the block may be used as a peripheral block. In such a configuration, the correction processing is targeted for the feature amount of the block in the window set around the block, and the same processing can be performed for any block in the image. Thereby, the correction process can be simplified.

  In this case, it is more preferable that the size of the rectangle can be changed. More specifically, for example, the second step may be executed in response to an input for setting information for determining the size of the rectangle. By making it possible to change the range of the peripheral blocks used for the correction process according to the appearance of defects, it is possible to further enhance the correction effect.

  Further, the feature amount may be expressed by a continuous function with the pixel value of each pixel constituting the image as a variable. In the feature quantity that satisfies such a condition, correction based on the value of the feature quantity of the peripheral block functions particularly effectively.

  For example, each of the blocks may be one of the pixels constituting the image. According to such a configuration, it is possible to make the correction function effectively for an image including a fine texture expressed in pixel units.

  The present invention can be applied to a technique for dividing an image into a plurality of regions in accordance with a feature value, and an imaging target in an image to be processed is arbitrary.

DESCRIPTION OF SYMBOLS 1 Imaging device 13 Imaging part 14 Control part 141 CPU
147 Input reception unit 148 Display unit C cells and the like M medium S105 to S106 First step S107 Second step S108 Third step W well

Claims (11)

A first step for obtaining a feature value for each block obtained by dividing an image into a plurality of blocks;
A second step of correcting the feature value for each block based on the feature value values of a plurality of peripheral blocks around the block;
An image processing method comprising: a third step of dividing the image into a plurality of regions in accordance with the feature value of each block after correction.   In the second step, the value of the feature value for each block is set to a first average value that is an average value of feature values of each of the peripheral blocks of the block, or each of the block and the peripheral blocks of the block. The image processing method according to claim 1, wherein correction is performed based on a second average value that is an average value of feature amounts.   The feature value of the block after correction is a value corresponding to a difference between the feature value of the block and the first average value or the second average value in the second step. The image processing method as described.   In the second step, the value of the block after correction is represented by a value expressed by a linear function having the difference value between the block feature value and the first average value or the second average value as a variable. The image processing method according to claim 3, wherein an amount value is used.   The image processing method according to claim 1, wherein the plurality of blocks have the same size.   The image according to any one of claims 1 to 5, wherein in the second step, the blocks other than the block, which are included in a rectangle having a predetermined size including the plurality of blocks and are centered on the block, are used as the peripheral blocks. Processing method.   The image processing method according to claim 6, wherein the size of the rectangle can be changed.   The image processing method according to claim 7, wherein the second step is executed in response to an input of setting information for determining the size of the rectangle.   The image processing method according to claim 1, wherein the feature amount is represented by a continuous function having a pixel value of each pixel constituting the image as a variable.   The image processing method according to claim 1, wherein each of the blocks is one of pixels constituting the image.   A control program for causing a computer to execute each step of the image processing method according to claim 1.