JP2008003793A - Image processing apparatus and image processing program - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide an image processing apparatus and an image processing program capable of extracting only required minimum feature values in feature extraction of a contour. <P>SOLUTION: In the case of extracting feature values of the contour on an image, index numbers are successively put to the contour of an object. Gradients with respect to the contour of the object are found out and then a plurality of identical gradient points are found out. Then distances (invariants) from a gradient (end line) located on the most end to respective other gradients are found out. Then, a plurality of points whose distances between the end line and gradients passing the identical gradient points are near are selected from the plurality of identical gradient points and an index number corresponding to the center value of the identical gradient points is stored. Then, a point where a difference between the index number corresponding to the identical gradient point and the index number corresponding to the center value is a threshold or less is selected from the selected identical gradient points, the average distance of distances between respective gradients passing these identical gradient points and the end line is found out and the average distance is set as one invariant. <P>COPYRIGHT: (C)2008,JPO&INPIT

Description

  The present invention relates to an image processing apparatus and an image processing program for extracting an outline of an object from an image and extracting features of the outline.

As a conventional image processing apparatus, for example, as described in Non-Patent Document 1, the contour of an object is extracted from an image, the same gradient point of the contour is obtained, and an invariant (feature) is obtained using the same gradient point. Is known to generate an object and recognize an object.
Amit Sethi, et al. “'Curve and Surface Duals and the Recognition of Curved 3D Objects from their Silhouettes”', International Journal of ComputerVision, vol. 58, no. 1, pp. 73-86, June 2004.

  However, in the above-described prior art, when the contour of an object is extracted from an image, the straight line portion of the contour may become a small wave shape due to the influence of noise, variations in illumination conditions, and the like. In this case, since many same gradient points exist in that portion, the number of dimensions of the generated feature quantity increases unnecessarily. Therefore, in subsequent processing, when the object is recognized based on the feature amount of the contour, not only the calculation amount increases, but also the recognition accuracy may be lowered.

  An object of the present invention is to provide an image processing apparatus and an image processing program capable of extracting only a minimum necessary feature amount in contour feature extraction.

  The present invention is an image processing apparatus that extracts a contour of an object from an image, obtains a plurality of the same gradient points of the contour, and extracts the features of the contour using the same gradient points. Of the feature quantity generating means for obtaining the quantity and the feature quantities of the plurality of identical gradient points obtained by the feature quantity generating means, the similarity of each feature quantity at any of the plurality of identical gradient points is higher than the first predetermined value A first determination unit that determines whether the proximity of each position at any of a plurality of identical gradient points is higher than a second predetermined value, and each feature by the first determination unit. When it is determined that the amount of similarity is higher than the first predetermined value and the proximity of each position is determined to be higher than the second predetermined value by the second determination means, Feature quantity integration means for integrating feature quantities into one feature quantity; It is characterized in further comprising.

  In such an image processing apparatus, when a contour of an object is extracted from an image and a noise or a change in illumination conditions occurs, a simple straight line portion existing in the contour becomes a small wave shape. The same gradient point exists. However, since these same gradient points are in the same straight line portion of the contour, the feature values of the same gradient points obtained by the feature value generation means are similar to each other, and the positions of the same gradient points are close to each other. . Therefore, in this case, it is determined by the first determination means that the similarity of the feature values of the same gradient points is higher than the first predetermined value, and the second determination means determines the proximity of the positions of the same gradient points. Since it is determined that the value is higher than the second predetermined value, the feature amount integration unit integrates the feature amounts of the same gradient points into one feature amount. As a result, an unnecessary increase in the number of dimensions of the feature amount is suppressed, and only the minimum necessary feature amount can be extracted.

  Preferably, the information processing apparatus further includes means for sequentially assigning index numbers on the contour, and the second determination means includes a distance between any plurality of identical gradient points and each index number corresponding to any plurality of identical gradient points. Based on this, it is determined whether the proximity of each position is higher than a second predetermined value. Since the index numbers are assigned in order on the contour, the closer the index number is, the shorter the distance between the same gradient points to which the index number is assigned. Accordingly, when the proximity of the positions of the same gradient points is determined by the second determination means, not only the distance between the same gradient points but also the index numbers corresponding to the same gradient points are used, for example, contours. Even when the shape is complicated, the proximity of the positions of the same gradient points can be obtained with high accuracy.

  Preferably, the feature amount generation means obtains feature amounts of a plurality of the same gradient points based on the distance between the gradients passing through the same gradient point. The distance between each gradient through the same gradient point is an invariant amount with respect to position, rotation and magnitude. By using such an invariant, when the object is recognized by performing contour matching processing after that, not only is the processing simpler than when pattern matching is performed on the entire contour, but the object is displayed on the image. Matching can be performed even if a part of the object is hidden.

  Further, the present invention is an image processing program for extracting a contour of an object from an image, obtaining a plurality of same gradient points of the contour, and causing a computer to execute image processing for extracting contour features using the same gradient points. The feature quantity generation procedure for obtaining the feature quantity of a plurality of the same gradient points, and the similarity of each feature quantity at any of the plurality of same gradient points among the feature quantities of the plurality of the same gradient points obtained by the feature quantity generation procedure A first determination procedure for determining whether or not the degree is higher than a first predetermined value, and a second determination procedure for determining whether or not the proximity of each position at any of a plurality of identical gradient points is higher than a second predetermined value When the first determination procedure determines that the similarity of each feature amount is higher than the first predetermined value and the second determination procedure determines that the proximity of each position is higher than the second predetermined value, One feature value at any of the same gradient points It is characterized in that to execute the feature quantity integration procedure to integrate the feature value to the computer.

  In the execution of such an image processing program, when noise or a change in illumination conditions occurs when extracting the contour of an object from an image, a simple straight line portion existing in the contour becomes a small wave shape. There are a plurality of the same gradient points. However, since these same gradient points are in the same straight line portion of the contour, the feature values of the same gradient points obtained in the feature value generation procedure are similar to each other, and the positions of the same gradient points are close to each other. . Therefore, in this case, it is determined in the first determination procedure that the similarity of the feature quantity of each same gradient point is higher than the first predetermined value, and in the second determination procedure, the proximity of the position of each same gradient point is determined. Since it is determined that the value is higher than the second predetermined value, the feature values at the same gradient point are integrated into one feature value in the feature value integration procedure. As a result, an unnecessary increase in the number of dimensions of the feature amount is suppressed, and only the minimum necessary feature amount can be extracted.

  Preferably, the computer further executes a procedure of sequentially assigning index numbers on the contour, and in the second determination procedure, the distance between any plurality of the same gradient points and each index corresponding to any of the plurality of the same gradient points. Based on the number, it is determined whether the proximity of each position is higher than a second predetermined value. Since the index numbers are assigned in order on the contour, the closer the index number is, the shorter the distance between the same gradient points to which the index number is assigned. Accordingly, when determining the proximity of the positions of the same gradient points in the second determination procedure, not only the distance between the same gradient points but also the index numbers corresponding to the same gradient points are used, for example, contours. Even when the shape is complicated, the proximity of the positions of the same gradient points can be obtained with high accuracy.

  Preferably, in the feature amount generation procedure, feature amounts of a plurality of the same gradient points are obtained based on the distance between the gradients passing through the same gradient point. The distance between each gradient through the same gradient point is an invariant amount with respect to position, rotation and magnitude. By using such an invariant, when the object is recognized by performing contour matching processing after that, not only is the processing simpler than when pattern matching is performed on the entire contour, but the object is displayed on the image. Matching can be performed even if a part of the object is hidden.

  According to the present invention, it is possible to extract only the minimum necessary feature amount in the feature extraction of the contour of the object. This makes it possible to reduce the amount of calculation and improve the recognition accuracy when the object recognition process is performed thereafter.

  DESCRIPTION OF EXEMPLARY EMBODIMENTS Hereinafter, preferred embodiments of an image processing apparatus and an image processing program according to the present invention will be described in detail with reference to the drawings.

  FIG. 1 is a block diagram showing a configuration of an embodiment of an image processing apparatus according to the present invention. The image processing apparatus 1 of this embodiment is mounted on a robot (not shown) that grips a container with a handle such as a teapot or a mug as a recognition target object.

  In FIG. 1, an image processing apparatus 1 inputs cameras 2A and 2B that capture a recognition target object and images captured by these cameras 2A and 2B, performs predetermined image processing, and estimates the posture of the recognition target object. An image processing unit 3, a monitor unit 4 for displaying the processing result of the image processing unit 3, and a data storage unit 5 for accumulating and storing a database used for image processing by the image processing unit 3 are provided.

  The cameras 2A and 2B are, for example, CCD cameras, and are provided in both eyes (not shown) of the robot.

  The image processing unit 3 may be configured as dedicated hardware specialized for object recognition processing, or a general-purpose computer such as a personal computer having a CPU, a memory (storage medium), an input unit, and an output unit is used. The general-purpose computer may execute an image processing program as software. At this time, the image processing program is provided by a storage medium such as a CD-ROM, a DVD, or a ROM, or a semiconductor memory. The image processing program may be provided via a network as a computer data signal superimposed on a carrier wave.

  Further, the processing result of the image processing unit 3 is sent to the grip control processing unit 6. The grip control processing unit 6 controls a robot hand (not shown) so as to grip the recognition target object based on the posture of the recognition target object estimated by the image processing unit 3.

  FIG. 2 is a flowchart showing an outline of a processing procedure by the image processing unit 3. In the figure, first, images captured by the cameras 2A and 2B are acquired (procedure 11). An example of images captured by the cameras 2A and 2B is shown in FIG. 3A shows an image (left image) captured by the camera 2A disposed in the left eye part of the robot, and FIG. 3B shows an image captured by the camera 2B disposed in the right eye part of the robot. Right image).

  Subsequently, the left image and the right image acquired in the procedure 11 are divided into regions for each object whose density values can be regarded as uniform (procedure 12). For example, in the image shown in FIG. 3, a region R surrounded by a frame is a region including a recognition target object.

  Further, in parallel with the region dividing process in step 12, a three-dimensional image is restored from the two-dimensional left image and right image (step 13). This three-dimensional image restoration is performed by, for example, using the concept of binocular parallax by calculating the depth of the point from the position coordinates of the point in the left and right images and the distance between the cameras 2A and 2B. Do.

  Subsequently, based on the two-dimensional image divided in the procedure 12 and the three-dimensional image restored in the procedure 13, the contour of the object existing in the region R is extracted by edge detection or the like (procedure 14). For example, the outline B of the recognition target object in the left image shown in FIG. 3A is as shown in FIG.

  Subsequently, the feature amount of the outline B of the recognition target object obtained in the procedure 14 is extracted (procedure 15). This feature extraction process is performed for both the left image and the right image. Details of the feature extraction process will be described later.

  Subsequently, the degree of similarity with respect to the feature matching data is calculated by matching the feature quantity of the outline B of the recognition target object obtained in step 15 with the feature matching data stored as a database in the data storage unit 5 ( Procedure 16). As the feature matching data, a lot of data obtained from captured images when the recognition target object is imaged from every angle (viewpoint) by the camera is registered. This matching process is performed for both the left image and the right image. As a matching method, for example, DP (Dynamic Programming) matching or the like is employed.

  Subsequently, of the left image and the right image, an image having a higher similarity to the feature matching data is selected as a reference image (procedure 17). Then, in the selected reference image, the maximum similarity to the feature matching data is calculated (procedure 18). Then, the contour shape corresponding to the feature matching data having the maximum similarity is set as the initial posture position of the recognition target object (procedure 19).

  Subsequently, the posture of the recognition target object is roughly estimated using only the reference image (procedure 20). Then, the validity of the posture estimation is determined (procedure 21), and when it is determined that the posture estimation has converged correctly, the posture of the recognition target object is estimated in detail using both the left image and the right image. (Procedure 22). The posture estimation process according to the procedures 20 and 22 is performed by, for example, creating a contour DT (Distance Transform) map. Then, the validity of the posture estimation according to the procedure 22 is determined (procedure 23), and when it is determined that the posture estimation has correctly converged, the estimation result is sent to the grip control processing unit 6 and displayed on the monitor unit 4. (Procedure 24).

  On the other hand, when it is determined in steps 20 and 22 that the posture estimation is not properly converged, the recognition target object is obtained using other feature matching data obtained from a captured image obtained by capturing the recognition target object from another viewpoint. Pose estimation. For example, feature matching data having the next highest similarity in the selected reference image is selected, and the contour shape corresponding to the feature matching data having the similarity is set as a new initial posture position of the recognition target object. (Procedure 25). And said procedure 20-24 is performed.

  FIG. 5 shows details of the feature extraction processing procedure of the procedure 15 described above. In the figure, first, an index point C is set for the contour of the object extracted in the procedure 14 shown in FIG. 2, and an index number is assigned to the index point C (procedure 31). The index number serves as an index indicating the positional relationship of the same gradient point P (described later) extracted from the contour of the object.

  Specifically, as shown in FIG. 6, for each index point C set on the contour B of the object, index numbers (1, 2,..., N− 1, n). The index points C have about several hundred to thousands on the contour B, and are arranged at regular intervals, for example.

  Further, a gradient (tangential gradient) H is obtained with respect to the contour B of the object (procedure 32). Then, a plurality of the same gradient points P are obtained for each angle of the gradient H (procedure 33). Here, the same gradient point P is a point where the gradient H is the same on the contour B as shown in FIG. For this reason, the gradients H corresponding to a plurality of the same gradient points P on the contour B are parallel to each other.

Then, determine the distance d _n to the most gradients on the edge (edge lines) each from H ₀ in the other gradient H _n among the plurality of gradient H _(n = 1, 2 ...) (Step 34). The distance d _n is the distance along the direction perpendicular to the gradient H _n.

The distance d _n from such end line H ₀ to the gradient H _n, the position of the contour B, and an amount invariant to rotation and size. By using this invariant as the feature quantity of the same gradient point P, the matching process of the procedure 16 shown in FIG. 2 can be easily performed. Further, for example, unlike when pattern matching is performed on the entire contour, it is possible to perform accurate matching even when a part of the contour B of the recognition target object is hidden.

The calculation from the end line H ₀ of the distance d _n to the gradient H _n is to only one of the two gradient H located at both ends of the contour B may be carried out with Tansen H _0, the contour B Both of the two gradients H located at both ends of the line may be performed as the end line H ₀ . As the invariant, except for using the distance d _n between the end line H ₀ to the gradient H _n may be directly used, such as maximum distance and the difference between the distance d _n in these distances d _n .

Subsequently, a plurality of out of the same gradient point P, the same slope points as the distance d _n is close to equal to or between the gradient H _n through the end line H ₀ the same gradient point P existing on the contour B P _i, that is the difference between the distance _{d n} to select the same slope points _{P i} as smaller than the threshold value (Step 35). For example, in the case shown in FIG. 7, the same gradient point P on the gradients H ₁ and H ₃ is selected as the same gradient point P _i .

Subsequently, the variance σ _i ² of the coordinates (x _i , y _i ) of the selected same gradient point P _i is obtained (procedure 36). Here, when the coordinate values of the same gradient points P _i are X ₁ , X ₂ ,..., X _N and the average value of the coordinate values of the same gradient points P _i is M, the variance σ _i ² is It is expressed by a formula.

Further, the median value m of the coordinates (x _i , y _i ) of the selected same gradient point P _i is obtained, and the index number I _m corresponding to this median value _m is stored (procedure 37).

Subsequently, the same gradient point P _j whose distance from the median m is equal to or less than the threshold (σ _i * α) is determined from the same gradient points P _i selected in the procedure 35 (procedure 38). At this time, for example, a threshold value proportional to the standard deviation σ _i is used in order to change the threshold value depending on the size of the contour B, but a constant value not depending on the standard deviation σ _i may be used as the threshold value. When a constant value is used as the threshold value, the process of step 36 can be omitted.

Subsequently, it is from the same gradient point P _j obtained in Step 38, the following difference threshold between the index number I _m corresponding to the index number I _j and the median m corresponding to the same gradient point P _j same A gradient point _Pk is selected (procedure 39). As a threshold at this time, a value corresponding to the total number of index points or a constant value is used.

Subsequently, the average value of the distance d _n between the gradient H _n and Tansen H ₀ through each identical gradient point P _k selected (average distance), and sets the average distance to the Invariant (Step 40 ). In other words, instead of setting all the invariants of the distance d _n between the gradient H _n and Tansen H ₀ through each identical gradient point P _k, and one invariant to integrate them.

  Thereafter, when it is determined whether or not invariant integration is possible for all the same gradient points P existing in the contour B, the feature extraction process is terminated. Otherwise, the procedure returns to step 35, and the undetermined same gradient point P is determined. Steps 35-40 are repeatedly executed for (Step 41).

  Note that the timing for executing the index numbering process of the procedure 31 is not particularly limited to the above, and may be any time as long as the process of the procedure 37 is executed.

  In such a feature extraction process, the procedure 31 constitutes means (procedure) for sequentially assigning index numbers on the contour. Procedures 32 to 34 constitute feature quantity generation means (feature quantity generation procedure) for obtaining feature quantities of a plurality of the same gradient points. In step 35, it is determined whether or not the similarity of each feature value at any of the plurality of same gradient points is higher than the first predetermined value among the feature values of the plurality of same gradient points obtained by the feature value generation means. It constitutes first determination means (first determination procedure). Procedures 36 to 39 constitute second judging means (second judging procedure) for judging whether or not the proximity of each position at an arbitrary plurality of identical gradient points is higher than a second predetermined value. In step 40, the first determination means determines that the similarity of each feature amount is higher than the first predetermined value, and the second determination means determines that the proximity of each position is higher than the second predetermined value. And feature quantity integration means (feature quantity integration procedure) for integrating each feature quantity at any of a plurality of identical gradient points into one feature quantity.

By the way, when disturbances such as noise and fluctuations in illumination state occur when acquiring images captured by the cameras 2A and 2B, as shown in FIG. 8, a slight undulation occurs in the contour B of the object on the image. The straight line part of B may not be a complete straight line. In this case, many same gradient points P exist only in the straight line portion. In this case, for each gradient H through each identical gradient point P, when the determined distance between the end line H _0, which is one of the gradient H of the (invariant) respectively, as shown in the example of the following formula, very The number of invariants with close values increases. In the following formula, the numerical value 0 represents the invariant of the same gradient point P on the end line H ₀ , and the other numerical values represent the invariant of the same gradient point P on the gradient H other than the end line H _0. ing.
D = [0 179.95 180.00 180.01 180.02 180.10 ... 180.20] (A)

  When the number of invariant dimensions is unnecessarily increased in this way, the calculation amount increases and the calculation time increases when calculating the similarity to the feature matching data. Further, when recognizing the type of object from the contour B, the recognition accuracy may be reduced.

On the other hand, in the present embodiment, among the plurality of the same gradient points P existing on the contour B, the same gradient point P _i on the plurality of gradients H having the same or close distance to the end line H ₀ , that is, not similarity variables to select a high multiple of the same slope points P _i with (see step 34 and 35 of FIG. 5), the index number I corresponding to the distance and the same gradient points P _i between each same gradient points P _{i Using m} , a plurality of the same gradient points P _k having a high degree of proximity of the positions of the same gradient points P _i are selected (see procedures 36 to 39 in FIG. 5). Then, an average distance from the end line H ₀ to each gradient H passing through each same gradient point P _k is obtained, and this average distance is set as one invariant (see procedure 40 in FIG. 5). Therefore, a plurality of invariants in the above equation (A) are integrated into one invariant as shown in the following equation.
H = [0 180.00] (B)

As a comparison, in the contour B shown in FIG. 7, since the difference in the distance _{d n} between the end line _{H 0} until the gradient _H 1, _{H 3} is small, invariant for each the same gradient point P on the gradient _H 1, _{H 3} However, the same gradient points P on the gradients H ₁ and H ₃ are not close to each other but separated from each other. For this reason, all of the invariants of the same gradient point P are determined to be necessary for feature extraction, and are extracted without being integrated.

  As described above, according to the present embodiment, when the invariants (features) of a plurality of the same gradient points P existing on the contour B of the object are similar and the positions of the same gradient points P are close to each other, Since the invariants of the same gradient points P are integrated into one, even if some undulation occurs in the straight line portion of the contour B due to noise, fluctuations in illumination state, etc., the number of invariant dimensions increases unnecessarily. Only the minimum necessary invariants can be generated. Thereby, when calculating the similarity to the feature matching data in the subsequent matching process (see step 16 in FIG. 2), the calculation time is reduced because the calculation amount is small. In addition, when the posture estimation process (see steps 21 and 23 in FIG. 2) is performed, the object recognition accuracy can be improved.

  At this time, for example, when the shape of the contour B is complex, even if the distance between the same gradient points P is short, the positional relationship of the same gradient points P along the contour B may be far apart. However, by using the index numbers given in order along the contour B, it is possible to accurately determine whether or not the positions of the same gradient points P are close to each other even in such a case.

  Further, when the contour B is smoothed in order to remove the undulation existing in the straight line portion of the contour B, detailed information may be lost and it may be difficult to extract the features of the contour B. Then, it is possible to generate only the minimum necessary invariant without erasing the feature of the contour B.

The present invention is not limited to the above embodiment. For example, in the above embodiment, as obtained based the positional relationship between the same slope point P existing on the contour B, and each index number I _m corresponding to the distance and the same gradient point P between the same slope point P However, when the shape of the contour B of the recognition target object is relatively simple, the positional relationship between the same gradient points P may be obtained based only on the distance between the same gradient points P.

  In the above embodiment, the parameter related to the distance between the gradients H of the contour B is used as the feature amount of each same gradient point P. However, the feature amount is invariant to the position, rotation, and size of the contour B. As long as it is an invariant, the present invention is not particularly limited to this. For example, curvature, color information, a histogram, or the like may be used.

  Further, although the above embodiment is applied to a robot that grips an object, the image processing apparatus and the image processing program of the present invention recognize other objects, estimate the posture of the object, and the like. It is also applicable to.

1 is a block diagram showing a configuration of an embodiment of an image processing apparatus according to the present invention. It is a flowchart which shows the outline of the process sequence by the image process part shown in FIG. It is a figure which shows an example of the left image and right image which were imaged with the two cameras shown in FIG. It is a figure which shows the outline of the object extracted from the left image shown to Fig.3 (a). It is a flowchart which shows the detail of the procedure of the feature-value extraction process shown in FIG. It is a conceptual diagram which shows a mode that the index number was attached | subjected on the outline. It is a conceptual diagram which shows a mode that several incline points are calculated | required with respect to an outline, and the invariant (feature amount) of each same gradient point is extracted. It is a figure which shows an example of the several same gradient point which exists in the linear part of an outline.

Explanation of symbols

  DESCRIPTION OF SYMBOLS 1 ... Image processing apparatus, 3 ... Image processing part (Characteristic quantity production | generation means, 1st judgment means, 2nd judgment means, feature-value integration means).

Claims (6)

An image processing apparatus that extracts a contour of an object from an image, obtains a plurality of same gradient points of the contour, and extracts features of the contour using the same gradient points,
Feature quantity generating means for obtaining feature quantities of the plurality of same gradient points;
A first determination is made to determine whether or not the degree of similarity of each feature value at any of a plurality of same gradient points is higher than a first predetermined value among the feature values of a plurality of same gradient points obtained by the feature value generation means. Means,
Second determination means for determining whether the proximity of each position at the plurality of the same gradient points is higher than a second predetermined value;
The first determining means determines that the similarity of each feature amount is higher than the first predetermined value, and the second determining means determines that the proximity of each position is higher than the second predetermined value. Then, an image processing apparatus comprising: a feature amount integration unit that integrates each feature amount at the plurality of arbitrary same gradient points into one feature amount. Means for sequentially attaching an index number on the contour;
The second determination means determines the proximity of each position based on the distance between the plurality of the same gradient points and the index numbers corresponding to the plurality of the same gradient points. The image processing apparatus according to claim 1, wherein it is determined whether the value is higher than the value.   The image processing apparatus according to claim 1, wherein the feature amount generation unit obtains feature amounts of the plurality of same gradient points based on a distance between the gradients passing through the same gradient point. An image processing program for extracting a contour of an object from an image, obtaining a plurality of same gradient points of the contour, and causing a computer to execute image processing for extracting features of the contour using the same gradient points,
A feature quantity generation procedure for obtaining feature quantities of the plurality of same gradient points;
A first determination is made to determine whether or not the similarity of each feature value at any of a plurality of same gradient points is higher than a first predetermined value among the feature values of the same gradient points obtained in the feature value generation procedure. Procedure and
A second determination procedure for determining whether the proximity of each position at the plurality of identical gradient points is higher than a second predetermined value;
In the first determination procedure, it is determined that the similarity of each feature amount is higher than the first predetermined value, and in the second determination procedure, it is determined that the proximity of each position is higher than the second predetermined value. Then, the image processing program causing the computer to execute a feature amount integration procedure for integrating each feature amount at the plurality of the same gradient points into one feature amount. Causing the computer to further execute a procedure for sequentially assigning an index number on the contour;
In the second determination procedure, based on the distance between the plurality of the same gradient points and the index numbers corresponding to the plurality of the same gradient points, the proximity of each position is the second predetermined value. 5. The image processing program according to claim 4, wherein it is determined whether or not the value is higher than the value.   6. The image processing program according to claim 4, wherein, in the feature amount generation procedure, feature amounts of the plurality of same gradient points are obtained based on a distance between the gradients passing through the same gradient point.

Images