WO2005101313A1 - Template matching device - Google Patents

Abstract

A template matching device includes: extraction means for extracting a partial image signal for calculation from an input image signal while successively changing the extraction position; calculation means for calculating an index concerning similarity between the partial image signal and a known template signal each time the partial image signal is extracted by the extraction means; and specification means for specifying a matched position with the template signal in the input image signal according to the index concerning the similarity at a plurality of positions of the input image signal. The extraction means decides a shift amount from the current extraction position to the next extraction position considering the index concerning the similarity in the current extraction position calculated by the calculation means.

Description

 Technical Field of Template Matching Apparatus.

 The present invention relates to a template matching apparatus that performs template matching on an input image signal including image information or audio information. Background art

 Template matching is a process of searching for a partial image signal (target) that matches a known template signal from the input image signal, and specifying the position of the target (match position). In this process, the matching operation between the calculation partial image signal extracted from the input image signal and the known template signal is repeatedly performed while the extraction position of the calculation partial image signal is gradually moved. Then, by comparing the results of the matching operation at each position in the input image signal, the matching position is specified. For the matching operation, a well-known cross-correlation method, a sequential error test, or the like is used. +

Further, when the extraction position of the partial image signal for calculation is moved little by little, the movement is uniformly performed at a preset interval in the entire input image signal or within a predetermined range (for example, see Japanese Patent Application Laid-Open No. 2000-200). (See Japanese Patent Application Laid-Open No. 0-764444 and Japanese Patent Application Laid-Open No. 7-22076). That is, the extraction position of the calculation partial image signal is moved at a fixed interval set before starting the template matching. In the above-mentioned Japanese Patent Application Laid-Open No. 2000-76044 and Japanese Patent Application Laid-Open No. Hei 7-22076, first, coarse template matching is performed uniformly at a predetermined large moving interval. After narrowing down to a predetermined range including the matching position, it has been proposed to perform the same fine template matching at a predetermined small movement interval within the range to specify the matching position. ing.

 However, the above method of uniformly performing template matching at a preset moving interval has a problem that it is difficult to set the moving interval to an appropriate constant value. An appropriate constant value of the moving interval is a value that can accurately and efficiently identify a matching position in the input image signal in a short time, and differs depending on the input image signal and the template signal.

 By the way, if the set value of the movement interval is smaller than an appropriate fixed value, the matching position can be specified (or the predetermined range including the matching position can be narrowed down) accurately, but the number of repetitions of the matching calculation is increased, so that an unnecessary It takes time. Conversely, if the set value of the movement interval is larger than an appropriate fixed value, the number of repetitions of the matching operation is reduced, thereby shortening the time. However, specifying the matching position (or narrowing down a predetermined range including the matching position) is accurate. This is not always the case, and it may become impossible to perform pseudo-matching or matching. Disclosure of the invention

 SUMMARY OF THE INVENTION An object of the present invention is to provide a template matching apparatus capable of reliably identifying a matching position with a reduced time.

The template matching apparatus according to the present invention includes: an extracting unit that sequentially changes an extraction position from among input image signals to extract a partial image signal for calculation; and a known image processing unit that extracts the partial image signal extracted by the extracting unit. A first calculating means for calculating a similarity with the template signal; and comparing the similarity obtained by the first calculating means in magnitude, thereby obtaining a matching position with the template signal in the input image signal. And a second calculating means for obtaining an index relating to the similarity between the partial image signal and a known template signal each time the partial image signal is extracted by the extracting means, Further, taking into account the index regarding the similarity at the current extraction position obtained by the second calculation means, the current extraction position is shifted from the current extraction position to the next extraction position. A moving amount determining unit for determining a moving amount in the extracting unit; and a similarity between the partial image signal and a known template signal in the first calculating unit based on a calculation result by the second calculating unit. At least with the judgment means for judging whether or not to perform the calculation of

It is characterized by having one.

 ADVANTAGE OF THE INVENTION According to this invention, a matching position can be pinpointed efficiently by shortening time reliably. Brief Description of Drawings

 FIG. 1 is a schematic diagram of the observation device 10.

 FIG. 2 is a diagram illustrating an input image 21, a template image 22, a target 23, and a partial image 24 for calculation.

 FIG. 3 is a flowchart showing an overall processing procedure in the template matching of the first embodiment. FIG. 4 is a flowchart showing a processing procedure in step S5 of FIG. 5A and 5B illustrate the density value Amn at the position (m, n) of the template image 22 and the density value B at the position (m, n) of the partial image 24 for calculation, respectively. FIG.

 FIG. 6A and FIG. 6B are diagrams illustrating the magnitude change of the movement interval (movement amount M x) according to the positional relationship with the coincidence position (X I, Y 1).

 FIG. 7 is a flowchart showing a part of the processing procedure in the template matching of the second embodiment.

 8A and 8B are diagrams for explaining the movement of the extraction position (X2, Y2) by the processing of FIG. ―

 FIGS. 9A, 9B, and 9C are diagrams for explaining the matching calculation based on the density histogram.

FIG. 10 is a flowchart showing the template matching processing procedure of the present embodiment. It is one.

 FIG. 11A, FIG. 11B, and FIG. 11C are diagrams illustrating the relationship between the asymmetry of the distribution shape of the density histogram and the skewness. BEST MODE FOR CARRYING OUT THE INVENTION

 Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings. In the first to fourth embodiments, the template matching process is step-controlled based on the similarity determination. In the fifth embodiment, the template calculation processing is cut-controlled in the matching calculation processing based on the similarity determination. In the sixth embodiment, control is performed by combining the processing of the first to fourth embodiments and the processing of the fifth embodiment.

 (First Embodiment)

 Here, the template matching apparatus and method according to the first embodiment will be described using the observation apparatus 10 shown in FIG. 1 as an example. The observation device 10 includes a stage 11, an optical system 12, a camera 13, and an image processing unit 14. Stage 11 supports sample 1 OA. The optical system 12 forms an optical image of a local region of the sample 1OA. The camera 13 captures an optical image of the sample 1OA with an image sensor (not shown), and outputs an image signal to the image processing unit 14.

 When the image processing unit 14 captures an image signal from the camera 13, it converts the image signal into a digital image of predetermined bits (for example, 8 bits), and stores the digital image in a memory (not shown) as an input image. Then, template matching described later is performed on the input image. The sample 10A is, for example, a semiconductor wafer, a liquid crystal substrate, a printed circuit board, a biological specimen (for example, cells), or the like. Using the observation device 10, observation, inspection, and alignment of the sample 1OA are performed.

Next, template matching in the image processing unit 14 will be described. Template matching is performed, for example, by using a known template from the input image 21 shown in FIG. This is a process of searching for a target 23 that is a partial image that matches the late image 22 and specifying the position of the evening target 23 (hereinafter referred to as “match position (位置 ,, Y)”. In the above, the matching operation between the partial image 24 for operation extracted from the input image 21 and the template image 22 causes the extraction position (Χ ₂ , Υ ₂ ) of the partial image 24 for operation to be slightly reduced. The extraction position (Χ ₂ , Υ ₂ ) is moved from the upper left corner to the lower right corner of the input image 21 in the figure. Here, it is assumed that each scanning line is parallel to the X direction, and the X direction parallel to the scanning line is called “horizontal direction”, and the Υ direction perpendicular to the scanning line is called “vertical direction”. .

 Note that the input image 21 corresponds to “input image signal including image information”. The template image 22 corresponds to a “template signal including image information”. The calculation partial image 24 corresponds to “calculation partial image signal including image information”.

The template processing of the first embodiment is performed by the image processing unit 14 according to the procedure (steps S1 to S6) of the flowchart shown in FIG. Also, the processing of step S5 in FIG. 3 (that is, the movement processing of the extraction position (X ₂ , Y ₂ ) of the partial image 24 for calculation) is performed according to the procedure of the flowchart shown in FIG. 15). Further, in the template matching of the first embodiment, the moving interval of the extraction position (X ₂ , Y ₂ ) of the partial image for calculation 24 is set to an appropriate constant value before starting this template matching. There is no need to keep it. The processing shown in FIG. 3 can be started as soon as the input image 21 is captured without performing such preprocessing. In step S1 of FIG. 3, the image processing unit 14 extracts a partial image 24 for calculation from the input image 21. The first extraction is performed, for example, from the initial position of the upper left corner of the input image 21 (FIG. 2). The calculation partial image 24 has a rectangular shape with the same size (number of pixels) in both the vertical and horizontal directions as the template image 22. FIG. 2 shows the vertical and horizontal sizes (the number of pixels) of the template image 22 by Tx and Ty. Both the partial image 24 and the template image 22 are smaller than the input image 21 (the number of pixels is small). Next (Step S2), the image processing unit 14 performs a matching operation between the template image 22 and the partial image 24 for calculation extracted in Step S1. Here, a known cross-correlation method is used as an example of the matching operation. In this case, the matching operation is performed by calculating the density value at the position (m, n) of the partial image 24 shown in FIG. 5; ^^ and the density value A _{mn at} the same position (m, n) of the template image 22. This is performed according to the following equation (1).

7, A _mn — A) (B _mn — B)

A: Average value of A _mn 'B: Average value of B _mn

The result of the matching calculation is a normalized cross-correlation coefficient C (1 1 ≤ C ≤ 1), which corresponds to the similarity between the calculation partial image 24 and the template image 22. The similarity is one of the “index of similarity” between the partial image 24 and the template image 22. The tendency is that the larger the cross-correlation coefficient C corresponding to the degree of similarity (ie, closer to “1”), the higher the similarity between the partial image 24 and the template image 22, and the extraction position (X ₂ , In many cases, Y ₂ ) is close to the matching position (XX,).

When the above-described matching calculation is completed, the image processing unit 14 determines the similarity, that is, the cross-correlation coefficient C, which is the result of the matching calculation, in the next step S 3, at the extraction position (24 ₂ ,対 応 Save in the memory in the image processing unit 14 in association with ₂ ). After the processing of these steps S 1 to S 3, the processing is completed for one extraction position in the input image _{21 (X 2, Y 2)} . Then, when the extraction position (Χ ₂ , Υ ₂ ) of the calculation partial image 24 is moved to the next position, that is, when Step S4 is Yes, the process proceeds to Step S5.

The process of step S 5 is a process of moving the extraction position of the partial image 24 for computation (X _2, Y _2), carried out One was the procedure of the flowchart shown in FIG. 4 (step SI 1 to S 15) . First (step S11), the image processing unit 14 moves to the next position. It is determined whether the direction is the horizontal direction (the direction parallel to the scanning line). For this determination, the X coordinate of the current position in the input image 21 (that is, the position coordinate on the scanning line) is used. Then, unless the X coordinate of the current position corresponds to the right end of the input image 21 (that is, the end point of the scanning line), it is determined to be “horizontal” (S 11 is Yes).

When the extraction position (X ₂ , Y ₂ ) of the calculation partial image 24 is to be moved in the horizontal direction, the image processing unit 14 proceeds to the process of step S12, and executes the similarity (the mutual phase relationship) at the current position. Calculate the amount of movement。 from the current position to the next position, taking into account the number C). Specifically, the amount of movement Μχ is calculated by substituting the similarity (cross-correlation coefficient C) at the current position into the “similarity” in the following equation (2). The calculation of the movement amount Μχ is very simple. .

 Mx = floor [(Tx / 4) X (1_1 ^ Χsimilarity)]-(2)

Equation (2) includes the number of pixels テ ン プ レ ー ト in the horizontal direction of the template image 22 (see FIG. 2). This number of pixels Τχ is the same as the number of pixels in the horizontal direction of the partial image 24 for calculation. The -coefficient is an empirical value (for example, 1/2) defined in the range that satisfies 0 <≤1. ΠΟΟΓ means that the extraction position (Χ ₂ , Υ ₂ ) is moved in units of pixels of the input image 21 so that the decimal part is rounded down and converted to an integer.

The amount of movement 求 め obtained by equation (2) has a low similarity at the current position (the mutual correlation coefficient C) (that is, close to “0”), and a low similarity between the partial image 24 and the template image 22. However, the larger the extraction position (Χ ₂ , Υ ₂ ) of the partial image 24 is from the matching position (Χ, Υ,), the larger the size is often. The maximum value of the movement amount Μχ is when the similarity is “0”, and is the floor (Tx / 4). The minimum value of the movement amount Mx is when the similarity is “1”, and is expressed as floor [(Tx / 4) X (1−k,)]. If the movement amount Mx is determined to be "0 pixels", the actual movement amount is set to "1 pixel".

When the above calculation of the movement amount is completed, the image processing unit 14 sets the extraction position (, Y ₂ ) of the partial image 24 for calculation in the next step S 13 by the movement amount Μχ in the horizontal direction from the current position. Move to the next position away. Then, in step S1 in Fig. 3, Return to the processing and repeat the extraction of the partial image 24 for calculation (S 1) —matching operation with the template image 22 (S 2) —save result (S 3) at the new extraction position (X ₂ , Y ₂ ) .

By repeating the above steps S 1 → S 2 → S 3 → S 4 → S 5 (S 11 → S 12 → S 13) → S 1 ..., the extraction position (X ₂ , Y ₂ ) reaches the end point of the scanning line (i.e. the right end of the input image 21), after the extraction position (chi _2, the process of step S 1 to S 3 against the Upsilon ₂₎ is completed, move further to the next position In this case (34 is 63), the image processing unit 14 determines that the moving direction to the next position is “vertical direction (direction perpendicular to the scanning line)” in step S11 in FIG. Step S11 is No). '.

When moving the extraction position (X ₂ , Y ₂ ) of the calculation partial image 24 in the vertical direction, the image processing unit 14 proceeds to the process of step S14, and performs similarity at each position on the current scanning line. Calculate the amount of movement My from the current scan line to the next scan line, taking into account the (cross-correlation coefficient C). Specifically, the similarity (cross-correlation coefficient C) at each position on the current scan line is read from the memory, and the average value of the similarities is calculated by the following equation (3), To calculate the travel distance My. Each position on the current scan line is, for example, all positions on the current scan line (including the current position). The calculation of such a movement amount My is very simple.

My floor [(Ty / 4) X ( row mean of 1-k ₂ x similarity)] - (3) in the equation (3), includes vertical pixel number Ty of the template image 22 (see FIG. 2) It is. This number of pixels Ty is the same as the number of pixels in the vertical direction of the partial image 24 for calculation. The coefficient k ₂ is an empirical value (for example, 1/2) determined within a range satisfying 0 and k ₂ ≤l. The amount of movement My calculated by equation (3) is small for the average similarity (cross-correlation coefficient C) in the current scanning line (ie, close to “0”), The similarity between the minute image 24 and the template image 22 is low on average, and becomes larger as the current scan line is farther from the matching position (Χ, Υ,). The maximum value of the travel distance My is 2005/007487

9 This is the case where the average value of the similarity is “0”, which is ΠΟΟΓ (Ty / 4). The minimum value of the movement amount My is when the average value of the similarity is “1”, and is floor [(Ty / 4) X (1−1 k ₂ )]. If the movement amount My is determined to be "0 pixel", the actual movement amount is set to "1 pixel". ■

When the above calculation of the movement amount My is completed, the image processing unit 14 sets the extraction position ('Χ ₂ , Υ ₂ ) of the partial image 24 for calculation in the next step S15 vertically from the current scanning line. It is moved in the direction to the next scanning line separated by the moving amount My, and in the horizontal direction to the left end of the input image 21 (that is, the starting point of the next scanning line). Then, returning to the process of step S1 in FIG. 3, at a new scanning line extraction position (X ₂ , Y ₂ ), extraction of the calculation partial image 24 (S 1) —combination with the template image 22 Matching operation (S 2)-Save result (S 3) is repeated.

In this way, by repeating the processing of steps S 1 → S 2 → S 3 → S 4 → S 5 (S 11 → S 12 → S 13 or S 11 → S 14 → S 15) → S 1 ... While moving the extraction position (X ₂ , Y ₂ ) of the calculation partial image 24 along the direction of the raster scanning composed of a plurality of scanning lines, the calculation partial image 24 is moved at each position on the plurality of scanning lines. Can be sequentially extracted. Further, every time the partial image 24 for calculation is extracted, a matching operation with the template image 22 is performed, and the result is stored.

Then, the extraction position (Χ ₂ , Υ ₂ ) of the calculation partial image 24 reaches the end point of the last scanning line (that is, the lower right end of the input image 21), and the processing of steps S 1 to S 3 there is performed. Upon completion, the image processing unit 14 proceeds to the process in step S6 without moving to the next position (No in S4). In step S6, the similarity (cross-correlation coefficient C) at a plurality of different positions in the input image 21 is compared in magnitude, whereby the value is the largest, and the similarity with the template image 22 is the highest. The extraction position (X ₂ , Y ₂ ) of the partial image 24 is specified as the matching position (Χ, Υ,). This completes the template matching process for the input image 21. ·

As described above, in the template matching of the first embodiment, TJP2005 / 007487

10 When the extraction position (X ₂ , Y ₂ ) of the image 24 is moved little by little, at least one of the similarities obtained up to now (including the similarity at the current position) is considered, and the equation ( 2) Or calculate the movement amount Mx, My from the current position to the next position by using equation (3) (see steps S12 and S14 in Fig. 4).

Therefore, for example, when the extraction position (X ₂ , Y ₂ ) is moved in the horizontal direction, if the similarity at the current position is small, the result of equation (2) (the amount of movement Μχ) becomes large. At positions away from Χ _1; Υ,), the extraction positions (Χ ₂ , Υ ₂ ) can be moved efficiently and appropriately at large intervals. Conversely, if the similarity at the current position is large, the result of equation (2) (the amount of movement 小 さ く) becomes small, and when the position approaches the matching position (Χ, Υ), the extraction position (Χ ₂ , Υ ₂ ) can be moved appropriately at small intervals. In particular, when it reaches the vicinity of the matching position (Χ ,, Υ,), the extraction position (Χ ₂ , Υ ₂ ) can be appropriately moved at an interval corresponding to “1 pixel”.

FIG. 6 (a) and FIG. 6 (b) show, for example, the magnitude change of the movement interval (movement amount Μχ) according to the positional relationship with the matching position (Χ, Υ, Υ). FIG 6 Alpha, extraction position (X _2, Y ₂₎ and the matching position _(Χ μ Υ,) a position step change in the relationship between the shown side by side in order of time. In FIG 6.alpha., Mutually extracted position adjacent in the longitudinal direction (Χ _2, Υ ₂₎ 'How the deviation amount of and corresponds to the moving distance (moving amount Μχ). Further, FIG 6.beta, extraction position (Χ _2, Υ ₂₎ the trajectory of the indicated "-". Also in these figures, it can be seen that the movement of the extraction position (Χ ₂ , Υ ₂ ) is large at a distance from the matching position (Χ, Υ,), and becomes smaller as the position approaches the matching position (XΥ).

On the other hand, the same is true when the extraction position (, Υ ₂ ) is moved in the vertical direction. If the average similarity at the current scanning line is small, the result of Equation (3) (the amount of movement My) increases. Therefore, the extraction position (Χ ₂ , Υ ₂ ) can be efficiently and appropriately moved at large intervals away from the matching position (Χ ^ Υ,). Conversely, if the average similarity at the current scan line is large, the result of equation (3) (the amount of movement My) will be small, and when the position approaches the match position (X _|; Y), the extraction position ( Χ ₂ , Υ ₂ ) are moved appropriately at small intervals. Can.

According to the template matching of the first embodiment, at least one of the similarities obtained up to the present (including the similarity at the current position) is fed back, and the matching position (the position away from X い) becomes larger. Since the extraction position (Χ ₂ , Υ ₂ ) of the partial image 24 is moved, the number of repetitions of the matching operation can be reduced, and the processing time of the template matching can be shortened. Since the extraction position (X ₂ , Y ₂ ) of the partial image 24 moves as the distance approaches (the minimum unit of the movement amount is “1 pixel”), the exact matching position (Χ ^, Υ,) Can be specified.

Therefore, according to the template matching of the first embodiment, the interval at which the extraction position (X ₂ , Y ₂ ) of the J partial image 24 is moved little by little as in the related art does not need to be set in advance to an appropriate constant value. During the template matching, the moving interval (moving amount Mx, My) is set according to the positional relationship with the matching position yes (Υ,). _Therefore , the matching position ( _Χι , _{Υ |} ) Can be accurately and efficiently specified in a short time. In addition, since there is no need for the difficult preprocessing of setting the movement interval to an appropriate constant value in advance as in the past, template matching (the processing in Fig. 3) can be started immediately after the input image 21 is captured. Is simplified. '

Furthermore, according to the template matching of the first embodiment, the extraction position (Χ ₂ , Υ ₂ ) of the calculation partial image 24 is moved along a predetermined direction (for example, the direction of raster scanning) and at least the current Considering the degree of similarity at the position (for example, using Equation (2) or Equation (3)), to calculate the movement amount Mx, My to the next position, the matching position ( _{移動 ι} , _Υι ) can easily set an appropriate movement interval according to the positional relationship. The calculation of the amount of movement Mx, My performed sequentially at each position is very simple, and the addition of this calculation does not complicate the template matching process.

Also, according to the template matching of the first embodiment, for example, equations (2) and (3) 7487

As shown in Fig. 12, when calculating the amount of movement M x and My to the next position, the size of the template image 22 (the number of pixels T x, T y: where the size of the template image is represented by the number of pixels) Therefore, the extraction position (X ₂ , Y ₂ ) of the calculation partial image 24 can be moved efficiently with reference to the size of the template image 22. Furthermore, according to the template matching of the first embodiment, since the maximum value of the movement amounts M x, My is set to 1 Z 4 of the size of the template image 22 (the number of pixels T x, T y), When the extraction position (X ₂ , Y ₂ ) of the partial image 24 for calculation is largely moved, it is possible to avoid a situation in which the partial image 24 passes through the evening get 23. Further, according to the template matching of the first embodiment, the extraction position (Χ ₂ , Υ ₂ ) of the calculation partial image 24 is moved along the raster scanning direction, and at the same time each position (上 のIn order to calculate the amount of movement My to the next scanning line in consideration of the similarity at all positions, for example, it is easy to match the position of the match (X,, Y) during template matching. An appropriate scanning line interval can be set.

 Further, according to the template matching of the first embodiment, the moving amount My to the next scanning line is calculated based on the average value of the similarities at each position (for example, all positions) on the current scanning line. (For example, using equation (3)), an appropriate scanning line interval can be set stably without being affected by noise.

 (Second embodiment)

In the first embodiment described above, when the extraction position (X ₂ , Y ₂ ) of the calculation partial image 24 is moved in the horizontal direction, the “similarity at the current position” The degree of movement 次 へ to the next position was calculated by taking only the degree into account. However, the present invention is not limited to this. In the second embodiment, in addition to the similarity at the current position, An example will be described below in which the “similarity at a position immediately before the current position” is also considered.

In the template matching of the second embodiment, the processing of steps S21 to S25 of the flowchart shown in FIG. 7 is performed instead of step S13 of the flowchart of FIG. 4 described above. The image processing unit 14 is used in step S 12 in FIG. 7487

After calculating the movement amount Mx to the next position in consideration of the similarity at the current position, the extraction position (X ₂ , Y ₂ ) of the partial image 24 for operation is actually moved. Then, the processing of steps S21 to S23 in FIG. 7 is performed.

In step S21, the image processing unit 14 determines whether or not the current position in the input image 21 corresponds to the starting point of the scanning line (that is, the left end of the input image 21). If it is located at the start point of the scanning line (that is, if S21 is Yes), the process proceeds to step S24, and the extraction position (X ₂ , Y ₂ ) of the partial image 24 for calculation is set in the horizontal direction. Move by the movement amount Mx (see Fig. 8A). After that, the process returns to step S1 in FIG.

On the other hand, if it is not located at the start point of the scanning line (that is, if S 21 is N 0), the process proceeds to step S 22, and based on the similarity (cross-correlation coefficient C) at the current position, It is determined whether or not conditional expression (4) is satisfied. The coefficient in the conditional expression (4) is the same as the coefficient in the above-mentioned expression (2). Threshold T _s, using the entire or al determined statistical value of the input image 21 (i.e. the average density and density range) is obtained by calculation by the following equation (5). .

(1—current similarity) <T _s- (4)

T _s = average concentration / concentration range (5)

If the condition (4) in step S22 is satisfied, the cross-correlation coefficient C corresponding to the current similarity is large (that is, close to “1”), and the partial image extracted at the current position 24 And the template image 22 has a high similarity, and the current extraction position (X ₂ , Y ₂ ) reaches the vicinity of the matching position (Χ, Υ,). Then, as a result of the determination in step S22, the similarity at the current position indicates that “the extraction position is ₂ , Υ ₂ ) has reached the vicinity of the matching position yes Υ,)” (that is, if S 22 If Yes, go to the next step S23.

In step S23, it is determined whether or not the following conditional expression (6) is satisfied based on the similarity at the current position and the “similarity at the position immediately before the current position”. Do. This In the conditional expression (6), the coefficient k ₃ is an empirical value determined within a range satisfying 0 <k ₃ ≤1. The threshold value T _B is calculated by the following equation (7) using statistical values (that is, standard deviation and density range) obtained from the entire input image 21. .

(Similarity of the current similarity one previous) X k ₃ > T _B- (6)

'T _B = standard deviation / concentration range_ (7)

The case where the conditional expression (6) in step S23 is satisfied is a case where the current similarity greatly increases as compared with the previous similarity. In other words, the previous similarity was extremely small compared to the current similarity. Therefore, the case where the conditional expression (6) is satisfied means that the cross-correlation coefficient C corresponding to the previous similarity is small (that is, close to “0”), and the part extracted at the previous position It is considered that the similarity between the image 24 and the template image 22 is low, and the previous extraction position (X ₂ , Y ₂ ) is far from the matching position (X, Y). When the result of the judgment indicates that the similarity at the previous position indicates that “the extracted position (Χ ₂ , Υ ₂ ) is far from the matching position Υ” (that is, if S 23 is Yes) Goes to the next step S25.

The case where the conditional expression (4) in step S22 is satisfied and the conditional expression (6) in step S23 is satisfied is that the extraction position (X ₂ , Y ₂ This is the case where the similarity sharply increases at the current position when the) is greatly moved. For example, a white point (maximum density value) included in the black background (minimum density value) of the input image 21 is This can happen with template matching like searching. When the conditional expressions (4) and (6) are satisfied, even if the extraction position (Χ ₂ , Υ ₂ ) is moved small considering the current large similarity, if the extraction position is moved forward as it is, the matching position will be ).

For this reason, in the processing of step S25, the next position is set between the current position and the immediately preceding position (see FIG. 8 Β), and the extraction position (Χ ₂ , Υ. ₂ ) is set to the immediately preceding position. After returning to the position, move it forward by the amount of movement Μχ. In other words, the previous (current position—current position) travel amount Μ X and the current (current position—next position) travel amount Μ X 05 007487

The extraction position (X ₂ , Y ₂ ) is moved backward according to the difference from 15. After that, the process returns to the process of step S1 in FIG.

In addition, when the above-mentioned conditional expression (4) is not satisfied (that is, when S22 is No), when the conditional expression (4) is satisfied but the conditional expression (6) is not satisfied (that is, S2 3 the cases of N o), step S 2 instead 5, proceeds to step S 2 4, thereby extracting position (X _2, Y ₂₎ is moved by the amount of movement M x forward. After that, the process returns to the process of step S1 in FIG.

Therefore, according to the template matching of the second embodiment, the partial image 24 for calculation passes through the evening get 23, that is, the extraction position (X ₂ , Y ₂ ) passes the matching position (X い). As a result, it is possible to reliably avoid a situation that may occur when the template matching is performed, so that the matching position (確 実 ,, Υ,) in the input image 21 can be reliably specified. As in the template matching of the first embodiment described above, it is needless to say that the matching position (χ, γ) can be specified accurately and efficiently in a short time.

 (3rd embodiment;) +

 In the first embodiment described above, an example is described in which the matching operation (see equation (1)) is performed by the cross-correlation method in step S2 of FIG. 3, but the present invention is not limited to this. In the embodiment, an example will be described in which a matching operation is performed by a residual sequential test method (SSDA method).

The matching operation by the residual sequential test method is based on the density value B at the position (m, n) of the partial image ₂₄ shown in FIG. 5 and the density value _{テ ン プ レ ー ト ιη at} the same position (m, n) of the template image _22. This is performed by the following equation (8) or equation (9), where D is the sum of absolute values of the density differences, and E in equation (9) is the sum of squares of the density differences. The results D and E of the matching operation correspond to the degree of similarity between the partial image 24 and the template image 22. The smaller the value (that is, the residual), the more similar the partial image 24 and the template image 22 are. This means that the extraction position (X ₂ , Y ₂ ) is close to the matching position (X ,, Υ,). Since the amount of calculation is small in the matching operation by the method, further speedup is possible.

 ^> =},? , l Amn-B mn l-(8)

 m n

E =), '. (Amn-B mn) ² … (9) When the matching operation by the residual sequential test method is performed in step S2 in FIG. 3, the movement amount in steps S12 and S14 in FIG. The following equations (10) and (11) may be used to calculate Mx and My. ,

 Mx = floor [(Tx / 4) X (k, x similarity)] — (10)

My = floor [(Ty / 4) X (k ₂ x average of similarity)] · '· (1 1) When calculating 移動, the similarity at the current position (result of matching operation D, E) is normalized, and the normalized value (0 to: L) is substituted into the “similarity” of equation (10). When calculating the movement amount My, the similarity at each position on the current scan line (results of the matching operation D, E) is normalized, and the average value of the normalized values (0 to 1) is calculated by the formula ( Substitute it into "Similarity row average" in 11). By such a calculation, an appropriate moving interval (moving amount Mx, My) can be easily set according to the positional relationship with the matching position (Υ,) during template matching, as described above.

Furthermore, when the matching operation by the residual sequential test method is performed in step S2 in FIG. 3, the determination in steps S22 and S23 in FIG. 7 (that is, the extraction position (X ₂ , Y ₂ ) The following conditional expressions (12) and (13) may be used to determine whether or not to return to.

(k, X current similarity) ku T _s — (12)

(Previous similarity-Current similarity) X k ₃ > T _B :-( 13)

In this case as well, the similarity (result of matching operation D, Ε) is normalized, and the normalized value (0 to 1) is substituted into conditional expressions (12) and (13). By the way, conditional expression (12) is used instead of conditional expression (4), and conditional expression (13) is replaced by conditional expression (6). Use instead. Threshold T _s of the condition (12) is the same as the formula (5). Threshold T _B of the condition (1.3) is the same as the formula (7). By using such conditional expressions (12) and (13), the partial image 24 for the calculation passes through the target 23 in the same manner as described above (that is, the extraction position (X ₂ , Y ₂ ) The position (by passing X ,, Υ) can be reliably avoided.

 (Fourth embodiment)

 The matching operation in step S2 in FIG. 3 is not limited to the cross-correlation method and the residual sequential test method described above, but may be a feature amount such as a density histogram or a moment of inertia. In the fourth embodiment, an example will be described in which a matching calculation is performed based on a density histogram and an inertia moment.

In the matching operation based on the density histogram, a density histogram H of the template image 22 is created in advance, for example, as shown in FIG. 9Α, and each time a partial image 24 for calculation is extracted, the density histogram H of the partial image 24 is extracted. H ₂₄ is created. The horizontal axis concentration values of concentration histograms H _24, H _22, the vertical axis is the number of pixels.

Then, the matching calculation is performed by counting the number of pixels in the overlapping portion of the two density histograms H ₂₄ and H ₂₂ (the hatched portion in FIG. 9B). The aggregate value F as a result of the matching operation in this case corresponds to the similarity between the partial image 24 and the template image 22. The larger the value, the higher the similarity between the partial image 24 and the template image 22. It indicates that the extraction position (X ₂ , Y ₂ ) is close to the matching position (Χ ,, Υ,).

 The moment of inertia is the distance (xi, yi) from the origin of all pixels of a certain area in the image and the density value wi to the inertia around the origin, as shown in the following equations (14) to (16). This is the value assigned to the moment formula. ·

Moment of inertia around the origin = ∑ (xiHyi ² ) wi… (14) Moment of inertia around each axis = ∑xi ² · wi or · '· (15)

Moment of inertia around each axis ∑yi ² 'wi… (16) 2005/007487

18 the matching operation based on the moment of inertia, for example created inertia 1 ₂₂ of the template image 22 in advance, for each extracted partial image 24 power for operation, the moment of inertia 1 ₂₄ of the partial image 24 is created. The absolute value of the difference between the two moments of inertia I ₂₄ and l _n is calculated, and the smaller the value is, the higher the similarity between the partial image 24 and the template image 22 is, and the extraction positions (X ₂ , Y ₂ ) match. Indicates that it is close to the position (Χ ,, Υ,).

 When the matching operation based on the density histogram and the matching operation based on the moment of inertia are performed in step S2 in FIG. 3 as described in FIGS. 9Α and 9 9, the movement amount Mx in steps S12 and S14 in FIG. , My can be calculated using the above equations (2) and (3).

 When calculating the movement amount Mx, the total value F, which is the similarity at the current position, is normalized, and the normalized value (0 to 1) is substituted for the “similarity” in equation (2). When calculating the movement amount My, the totalized value F, which is the similarity at each position on the current scan line, is normalized, and the average value of the normalized values (0 to 1) is calculated as "" in equation (3). Substituting into "Line mean of similarity". By such a calculation, an appropriate moving interval (moving amount Mx, My) can be easily set according to the positional relationship with the matching position (X ,, Υ,) during template matching as described above. .

Also, as described in FIGS. 9A and 9B, when the matching calculation based on the density histogram is performed in step S2 in FIG. 3, the determination in steps S22 and S23 in FIG. X ₂ , Y ₂ ) can be determined using the above conditional expressions (4) and (6).

In this case, as in the above case, the similarity (total value F) is normalized, and the normalized value (0 to 1) is substituted into conditional expressions (4) and (6). In this case, as described above, parts of image 24 for computation will past the target 23 (i.e. extraction position (chi _2, Upsilon ₂₎ will past means to match the position (chi ,, Upsilon,)) Such a situation can be reliably avoided.

Furthermore, the matching operation based on the density histogram is performed, for example, as shown in Fig. 9Α. This may be performed by counting the number of pixels of the absolute value of the difference between the two density histograms H ₂₄ and H ₂₂ (the hatched portion in FIG. 9C). The aggregate value G, which is the result of the matching operation in this case, corresponds to the similarity between the partial image 24 and the template image 22, and the smaller the value, the higher the similarity between the partial image 24 and the template image 22. It indicates that the extraction position (X ₂ , Υ,) is close to the matching position (X Υ).

 As described in Fig. 9Α and Fig. 9C, when the matching calculation using the density histogram is performed in step S2 in Fig. 3, the calculation of the movement amounts Mx and My in steps S12 and S14 in Fig. 4 The above equations (10) and (11) may be used.

 When calculating the amount of movement, the similarity (total value G) at the current position is normalized, and the normalized value (0 to 1) is substituted for the “similarity” in equation (10). When calculating the movement amount My, the similarity (total value G) at each position on the current scanning line is normalized, and the average value of the normalized values (0 to 1) is calculated as " Substituting into "Line mean of similarity". By such a calculation, an appropriate moving interval (moving amount Mx, My) can be easily set according to the positional relationship with the matching position (X ,, Υ,) during template matching, as described above.

Further, as described in FIGS. 9A and 9C, when the matching calculation based on the density histogram is performed in step S2 in FIG. 3, the determination in steps S22 and S23 in FIG. X ₂ , Y ₂ ) can be determined using the above conditional expressions (1 2) and (1 3). However, the threshold value Τ _の of the conditional expression (13) is calculated by the following expression (17).

Τ _Β = (standard deviation) Χ 2 / concentration range （(17)

In this case as well, the similarity (total value G) is normalized, and the normalized values (0 to 1) are substituted into conditional expressions (12) and (13). Therefore, as in the above, it is possible to reliably avoid the situation in which the partial image 2 for calculation passes the target 23 (that is, the extraction position (Χ ₂ , 位置₂ ) passes the ,, Υ,) at the matching position). it can.

In the first to fourth embodiments, the extraction position (位置₂ , When moving Y ₂ ) little by little, the size (number of pixels T x, Ty) of the template image 22 is taken into account, and the size corresponding to 1Z4 is taken as the maximum value of the movement amount Mx, My. Is not limited to this. The maximum values of the movement amounts Mx and My may be set to values other than “1Z4 of the number of pixels Tx and 'Ty” (for example, 1 × 2 of the number of pixels Tx and Ty). The setting of the maximum value of the movement amount Mx, My may be changed according to the type of the matching operation. For example, when performing a matching operation based on the density histogram, the maximum values of the movement amounts MX and My can be set larger than those of the cross-correlation method and the residual sequential test method. Therefore, it is possible to further increase the speed.

 In the above embodiment, when calculating the movement amount My of the raster scan to the next scan line, the “index of similarity” at each position (for example, all positions) on the current scan line is considered. However, the present invention is not limited to this. The present invention can be applied not only to all positions but also to a case where a part (for example, every other position) on the current scanning line is selectively considered. In that case, the "similarity indicator" at the current position need not be considered.

 Further, in the above embodiment, when calculating the movement amount My of the raster scan to the next scan line, the average “similarity” at each position (at least a part of positions) on the current scan line is calculated. However, the present invention is not limited to this. The present invention is applicable not only to the average value but also to the case where the largest “index of similarity” at each position (all or part) on the current scanning line is considered.

Further, in the above-described embodiment, an example has been described in which the extraction position (X ₂ , Y ₂ ) of the partial image 24 for calculation is moved along the raster scanning direction, but the present invention is not limited to this. . When moving the extraction position (Χ ₂ , Υ ₂ ) along another predetermined direction (for example, one-dimensional direction), the same calculation formula for moving amount (for example, formula (2)) as above is used. By doing so, a similar effect can be obtained.

Further, in the above-described embodiment, an example in which the moving direction of the extraction position (Χ ₂ , Υ ₂ ) is determined in advance (for example, an example of moving along the raster scanning direction) has been described. It is not limited to this. The moving direction of the extraction position (χ ₂ , Υ ₂ ) is determined in advance without considering the moving direction of the extraction position (X ₂ , Y ₂ ) and considering at least one of the “similarity indicators” obtained so far. Both the direction and the movement amount may be determined. In this case, the "similarity indicator" at the current position need not be considered.

 (Fifth embodiment)

 In the first to fourth embodiments, the template matching processing is step-controlled based on the similarity determination. In the fifth embodiment, the matching processing is cut off based on the similarity determination in the template matching processing. Control.

 The template matching according to the present embodiment is performed according to the procedure (steps S1 to S8) of the flowchart shown in FIG.

In the processing of FIG. 10, the matching operation (S 4) between the operation partial image 24 extracted from the input image 21 and the template image 22 is performed to extract the operation partial image 24. It is repeated as needed while moving the position (X ₂ , Y ₂ ) little by little. The input image 21 corresponds to “input signal including image information”. The template image 22 corresponds to a “template signal including image information”. The calculation partial image 24 corresponds to “calculation partial signal including image information”.

In the template matching according to the present embodiment, the skewness S22 of the template image ₂₂ is calculated up to the first time of the determination processing (S3) of whether or not to perform the matching operation (S4). Is a statistical scale (skewness) indicating the asymmetry of the distribution shape of the density histogram of the template image _22. Specifically, “X _ran ” in the following equation (18) Then, the skewness S22 of the template image ₂₂ is calculated by substituting the density value A „ _ra at the position (m,) of the template image ₂₂ shown in Fig. 5A. N is the total number of pixels in the template image 22, and is the standard deviation.

^{S = TO}ゝ⁽⁸⁾

X: The skewness S ₂₂ of the mean value template image 22 of Xmn, is described with reference to concentration histograms shown in FIGS. 11A-11 C. The horizontal axis of the density histogram shown in Fig. 11A-Fig. 11C is the density value, and the vertical axis is the number of pixels. When the distribution shape is symmetric as in the density histogram shown in FIG. 11A, the skewness S ₂₂ = 0. When the distribution shape is asymmetric and skewed to the left (that is, there are many pixels with low density) as in the density histogram shown in FIG. 1 IB, the skewness S ₂₂ <0. When the distribution shape is asymmetric and skewed to the right (that is, many pixels with high density) as in the density histogram shown in FIG. 11C, the skewness S ₂₂ > 0. Calculation of skewness S ₂₂ of the template image 22, as described above, is performed until the first pine quenching operation (S 4) performs judging whether processing of (S 3). The processing (S1 to S8) in FIG. 10 will be described in order.

 In step S1, the image processing unit 14 extracts a partial image 24 for calculation from the input image 21. The calculation partial image 24 has a rectangular shape having the same size (number of pixels) in both the vertical and horizontal directions as the template image 22. The calculation partial image 24 and the template image 22 are both smaller than the input image 21 (the number of pixels is small).

Next (Step S 2), the image processing unit 14 calculates the skewness S ₂₄ of the partial image 24 for computation extracted in step S 1. Skewness S ₂₄ in this case is a statistical measure of the asymmetry of the concentration histograms of the distribution shape of the partial image 24 (skeraess). Specifically, by substituting the density value B _mn at the position (m, n) of the partial image 24 shown in FIG. 5B into “X” of the above equation (1), the distortion of the partial image 24 for calculation is obtained. The calculation of the degree S ₂₄ is performed, where N in Equation (1) is the total number of pixels of the partial image 24, and σ is the standard deviation.

Skewness S ₂₄ of the partial image 24 is also similar to the skewness S ₂₂ of the template image 22, FIG. 11 When the distribution shape is symmetric as in the density histogram shown in A, the skewness S _{24 is} zero. If the distribution shape is asymmetric and skewed to the left (many pixels with low density) as in the density histogram shown in FIG. 11B, the skewness S ₂₄ <0. When the distribution shape is asymmetric and skewed to the right (many pixels with high density) as in the density histogram (c) shown in Fig. 11C, the skewness S ₂₄ > 0.

Next (Step S 3), the image processing unit 1 4, Ri by the comparison of the skewness S _M already computed the template image 22 and skewness ₂₄ of the partial image 24 calculated in step S 2, the skewness Calculate the absolute difference IS ₂₂ — S ₂₄ I. The absolute difference in skewness IS ₂₂ —S ₂₄ I is an index relating to the similarity between the density histogram of the template image ₂₂ and the density histogram of the partial image 24. The higher the similarity of the two density histograms, the smaller the absolute difference in skewness IS ₂₂ — S ₂₄ I. Conversely, the lower the similarity between the two density histograms, the greater the absolute skewness difference IS ₂₂ — S ₂₄ |.

Furthermore, when the absolute difference in skewness IS ₂₂ — S ₂₄ I shows a large value and the similarity of the two density histograms is low, the similarity between the template image ₂₂ and the partial image 24 is clearly low (the difference it seems to do. Conversely, if the absolute difference in skewness IS ₂₂ — S ₂₄ l shows a small value and the two density histograms have a high similarity, the similarity between the template image ₂₂ and the partial image 24 may be high. is there. In other words, similarity may be low, but similarity may be high. Therefore, the absolute difference in skewness IS ₂₂ —S ₂₄ I is considered to be an approximate value regarding the similarity between the template image 22 and the partial image 24.

After calculating the absolute difference in skewness IS ₂₂ _S ₂₄ |, the image processing unit 14 compares the absolute value with a predetermined threshold α (> 0). Then, it is determined whether or not the following conditional expression (19) is satisfied (step S3).

IS _n -S _u \ <ー (1 9)

When conditional expression (19) is not satisfied (No in S3), the similarity between template image 22 and partial image 24 is clearly low (different) as described above. It is not necessary to determine an accurate index (that is, similarity) regarding the similarity between the partial image 24 and the template image 22 by the log operation (S4). Therefore, when the conditional expression (19) is not satisfied (S3 is No), the matching operation (S4) is omitted, and the process proceeds to step S6 described later.

 On the other hand, when the conditional expression (19) is satisfied (that is, when S 3 is Yes), the similarity between the template image 22 and the partial image 24 may be high as described above, so that the matching operation is performed. By (S4), it is meaningful to obtain an accurate index (that is, similarity) regarding the similarity between the partial image 24 and the template image 22. Therefore, when the conditional expression (19) is satisfied (S3 is Yes), the process proceeds to step S4, and the matching operation (S4) is executed.

As described above, in the template matching (FIG. 10) of the present embodiment, in step S3, the approximate value (that is, the absolute difference in skewness IS ₂₂ ) relating to the similarity between the calculation partial image 24 and the template image 22 is obtained. - based on the S ₂₄ l), it determines whether or not to perform the matching operation between the partial image 24 and the template image 2 2 (S 4). The determination process of step S3 is a process of determining the similarity between the partial image 24 and the template image 22 roughly. Therefore, this calculation process can be performed in a shorter time than the matching calculation. In other words, the calculation speed is high. Then, when it is determined that the matching calculation is not performed, the calculation is omitted, and only when it is determined that the matching calculation is performed, the calculation is performed.

If the conditional expression (19) in step S3 is satisfied, the image processing unit 14 performs a matching operation (operation of similarity) between the operation partial image 24 and the template image 22 in step S4. . Here, a known cross-correlation method is used as an example of the matching operation. In this case, the matching calculation is performed based on the density value B _{M at the} position (m, n) of the partial image 24 shown in FIG. 5B and the density value at the same position (m, n) of the template image 22 shown in FIG. 5A. using the a _m, it is performed by the following equation (2 0). _, _, (Amn- A) (B _mn -1 ")

A: Average value of Amn 'B: Average value of B _mn

The result of the matching operation is a normalized cross-correlation coefficient C (1 1 C ≤ 1), which corresponds to the similarity between the operation partial image 24 and the template image 22. The similarity is one of the “index of similarity” between the partial image 24 and the template image 22, and is an “exact index of similarity”. The tendency is that the larger the mutual correlation number C corresponding to the degree of similarity is (the closer to “1”), the higher the similarity between the partial image 24 and the template image 22 is, and the extraction position (X ₂ , Y ₂ ) is close to the matching position (Χ ,, Υ,).

When the above-described matching calculation is completed, the image processing unit 14 determines the similarity (cross-correlation coefficient C), which is the result of the matching calculation, in the next step S5 by using the extraction position (Χ ₂ , Υ ₂ ) and store them in memory. After the processing of these steps S 1~ S 5, 1 single extraction position (Χ _2, Υ ₂₎ in the input image 21 processing for ends. When the moving extraction position of the partial image 24 for computation (chi _2, Upsilon ₂₎ the next position (i.e., when the step S 6 is Ye s), the processing proceeds to step S 7.

The process of step S7 is a process of moving the extraction position (X ₂ , Y ₂ ) of the partial image 24 for calculation. The image processing section 14, extraction position of the partial image 24 (Χ _2, Υ ₂₎ to move from the current position to the next position. Thereafter, the flow returns to the processing step S 1, in the new extraction position (X _2, Υ _2), extraction of the partial image 24 for computation (S 1) - calculation of the skewness S ₂₄ (S 2) - the matching operation Judgment of execution / omission (S3) → [Matching operation (S4)-Save result (S5) if necessary] is repeated.

In this way, by repeating the processing of steps S 1 → S 2 → S 3 → [S 4 → S 5] → S 6 → S 7 → S 1…, the extraction position (X ₂ , Y ₂ ) 2005/007487

 While moving 26 little by little, the partial images 24 are sequentially extracted at each position, and each time the partial images 24 are extracted, it can be determined whether or not to perform the matching operation (S 4). Only when the content of the judgment indicates “perform”, the matching operation with the template image 22 is performed, and the result is stored.

Then, when the extraction position (X ₂ , Y ₂ ) of the partial image 24 for operation reaches the end point, and the processing of steps S1 to S5 there ends, the image processing unit 14 moves to the next position. Without moving (S6 is No), the process proceeds to step S8. In step S8, the magnitude of the similarity (cross-correlation coefficient C) actually obtained at a plurality of different positions in the input image 21 is compared with each other to obtain the largest value. The extraction position (X ₂ , Y ₂ ) of the partial image 24 having the highest similarity is identified as the matching position (Χ, Υ,). This is the end of the template matching processing for the input image 21. -As described above, in the template matching of the present embodiment, each time a partial image 24 for calculation is extracted, it is determined whether or not to perform a matching calculation between the partial image 24 and the template image 22. If not, skip the matching operation. Then, the matching operation is performed only when it is determined to “perform”. For this reason, the matching position (Χ Υ,) can be specified efficiently with a reduced time. In other words, the processing speed of template matching is dramatically improved.

In the template matching according to the present embodiment, matching is performed based on an approximate value (ie, absolute difference in skewness IS ₂₂ —S ₂₄ I) regarding the similarity between the partial image ₂₄ for calculation and the template image _22. It is determined whether to perform the operation. Therefore, when the template image 22 and the partial image 24 are clearly different (that is, when the similarity is clearly low), unnecessary matching calculation can be efficiently omitted. The threshold value α of the conditional expression (19) is calculated by, for example, calculating the skewness (skewnes s) in each case where the contrast of the template image 22 is changed, and calculating the range value (= maximum value) of the obtained calculation result. It is preferable to set in consideration of (skewness-minimum skewness). JP2005 / 007487

 27

—As an example, it is conceivable to set the threshold value α == range value. Also, the threshold α may be set to the range value X coefficient k. The coefficient k in this case is an arbitrary value of 1/2 to 1. By the way, if the set value of the threshold value α is too large, the omission rate of the matching calculation decreases, and the improvement of the processing speed cannot be expected. Conversely, if the set value of the threshold value is too small, it is difficult to specify the exact matching position (Χ _1; Υ) due to the influence of the contrast change and noise of the partial image 24 for calculation.

In the fifth embodiment, whether or not to perform the matching operation by using the absolute difference of skewness IS ₂₂ —S ₂₄ I as an approximate value related to the similarity between the calculation partial image ₂₄ and the template image ₂₂ However, the present invention is not limited to this. Other operations may be used as long as the operation can be performed in a shorter time than the matching operation and the similarity can be determined.

In addition, the third moment M22 of the template image ₂₂ is calculated by the following equation (2 1). Similarly, the third moment M24 of the partial image ₂₄ for calculation is calculated, and these moments are calculated. M _22, M _M of absolute differences IM ₂₂ - using M ₂₄ I as an approximate value of the above may be determined whether or not to perform pine quenching operation. Also, not limited to the third order, an nth order (n is an odd number of 5 or more) moments may be used. The odd-order moments, like the skewness (skewnes s) already described, indicate the bias direction (left or right) of the density histogram, and are good indicators for determining whether or not to perform a matching operation. .

… (twenty one )

 mn In particular, when a third-order moment is used, the amount of calculation is smaller than that of skewness, so that the necessity of the matching operation can be determined at high speed. Therefore, even when the matching operation is performed by a method with a small amount of calculation (for example, a residual sequential test method described later), it is possible to reliably improve the processing speed.

Further, as another specific example of the above approximate values, the “absolute difference of the average density” obtained by comparing the average density of the template image 22 with the average density of the partial image 24 is used. You can also. In this case, the calculation amount is further reduced, and the above approximate value can be easily calculated. As a result, higher speeds are realized.

 Further, in the above-described embodiment, an example has been described in which the matching operation (see equation (20)) is performed by the cross-correlation method in step S4 of FIG. 10, but the present invention is not limited to this. In addition, the matching operation may be performed by a residual sequential test method (SSDA method), or the matching operation may be performed using a feature amount such as a density histogram or an inertia moment.

The matching operation by the residual sequential test method is based on the density value B _{m at} the position (m, n) of the partial image 24 shown in FIG. 5B and the density value at the same position (m, n) of the template image 22. This is performed by the following equation (2 2) or equation (2 3) using A _mn . D in equation (2 2) is the sum of absolute values of the density difference. E in equation (2 3) is the sum of squares of the concentration difference. The results D and E of the matching operation correspond to the similarity between the partial image 24 and the template image 22. The similarity is high, indicating that the extraction position (X ₂ , Y ₂ ) is close to the matching position (X い,). Since the matching operation by the residual sequential test method requires a small amount of calculation, it can be further speeded up. '

D = _mn -B _mn |… (22)

In the matching calculation based on the density histogram, for example, as shown in FIG. 9A, a density histogram H22 of the template image ₂₂ is created in advance, and each time a partial image ₂₄ for calculation is extracted, the partial image 2 A density histogram H _{M of} 4 is created. The horizontal axis of the density histograms H _M and H ₂₂ is the density value, and the vertical axis is the number of pixels.

Then, the matching calculation is performed by counting the number of pixels in the overlapping portion of the two density histograms H ₂₄ and H ₂₂ (the hatched portion in FIG. 9B). In this case, the result of the matching operation (total value F) is the similarity between the partial image 24 and the template image 22. Correspondingly, the larger the value, the higher the similarity between the partial image 24 and the template image 22, indicating that the extraction position (X ₂ , Y ₂ ) is closer to the matching position (, Υ).

Further, the matching calculation based on the density histogram may be performed by, for example, summing up the number of pixels of the absolute value of the difference between the two density histograms Η ₂₄ and Η ₂₂ shown in FIG. 9 (the hatched portion in FIG. 9C). . The result of the matching operation (total value G) in this case corresponds to the similarity between the partial image 24 and the template image 22. The smaller the value, the more similar the partial image 24 and the template image 22 are. Is high and the extraction position (Χ ₂ , Ύ ₂ ) is close to the matching position (Χ, Χ).

 (Sixth embodiment)

 The sixth embodiment of the present invention is a combination of the first to fourth embodiments and the fifth embodiment. The template matching apparatus of the sixth embodiment combines the cut control of the matching operation disclosed in the fifth embodiment with the step control of the template matching process disclosed in the first to fourth embodiments.

 More specifically, the template matching processing of the sixth embodiment performs the processing from step S1 to step S8 in FIG. 10 (matching calculation cut control). At that time, the step S1 in FIG. The “movement processing of the extraction position of the partial image for calculation” of 7 executes the processing (step control) from step SI1 to step S15 in FIG. As a result, the effects of both embodiments are synergistically achieved, and the speed of the template matching process is dramatically increased.

 Further, in the above-described first to sixth embodiments, an example has been described in which the shapes of the calculation partial image 24 and the template image 22 are rectangular, but the present invention is not limited to this. The partial image 24 and the template image 22 need not be rectangular as long as they have the same shape and size (number of pixels). The essential processing is the same for other shapes.

Furthermore, in Embodiments 1 to 6 described above, template matching has been described using an optical microscope apparatus as the observation apparatus 10 in FIG. 1 as an example, but the present invention is not limited to this. In addition, an electron microscope that scans a local area of the sample with an electron beam and captures images 5 007487

 The present invention can be applied to 30 devices. The present invention can be applied not only to the image of the local region of the sample but also to an apparatus which collectively captures the image of the entire surface of the sample. Similar effects can be obtained even when an external computer connected to the observation device 10 or the like is used. In the above-described first to sixth embodiments, examples have been described in which template matching is performed on an input image (input image signal including image information) captured from the camera 13 of the observation device 10 in FIG. However, the present invention is not limited to this. In addition, the present invention can be applied to a case where template matching is performed on an input image signal (input audio) including audio information (signal processing). When a one-dimensional input image signal such as input sound is targeted, the above-mentioned “image density” may be replaced with “signal strength”. “Average concentration” corresponds to “average intensity”.

Claims

The scope of the claims

1. extracting means for sequentially changing the extraction position from the input image signal to extract a partial image signal for calculation;

 5 calculating means for calculating an index regarding similarity between the partial image signal and a known template signal each time the partial image signal is extracted by the extracting means;

 Identifying means for identifying a matching position with the template signal in the input image signal, based on an index regarding the similarity at a plurality of positions of the input image signal.

 The extraction means determines a movement amount from the current extraction position to a next extraction position in consideration of the similarity index at the current extraction position obtained by the calculation means.

 'A template matching device characterized in that:

2. Extraction means for extracting a partial image signal for calculation by sequentially changing the extraction position from the input image signal;

 Calculating means for obtaining an index relating to the similarity between the partial image signal and a known template signal each time the partial image signal is extracted by the extracting means;

 Specifying means for specifying a matching position of the input image signal with the template signal based on the index regarding the similarity at a plurality of positions; Storage means for storing an index, wherein the extraction means considers at least one of the indices relating to the similarity, which is stored in the storage means, and which has been obtained by the arithmetic means up to now. A template matching device for determining a moving amount to an extraction position.

3. The template matching apparatus according to claim 2, The extraction means considers at least an index regarding the similarity at the current position, and considers the magnitude of the template signal, and calculates a movement amount from the current extraction position to the next extraction position. A template matching device characterized in that it is determined.

4. The template matching apparatus according to claim 1 or 2, wherein the extraction unit determines that the index regarding the similarity at the current position is “the extraction position has reached the vicinity of the matching position”. , And when the index regarding the similarity at the position immediately before the current position indicates that “the extracted position is distant from the matching position”, the next extracted position is Determine between the current extraction position and the previous extraction position

 A template matching apparatus characterized in that:

5. The template matching apparatus according to claim 1 or 2, wherein the input image signal, the partial image signal, and the template signal each include image information,

 The extracting means sequentially extracts the partial image signals while moving the extraction position along a raster scanning direction including a plurality of scanning lines, and also shifts the extraction position from a current scanning line to a next scanning line. At the time of moving, the movement amount from the current scanning line to the next scanning line is calculated in consideration of the similarity index at each position on the current scanning line, and based on the moving amount. A template matching device, wherein the extraction position is moved.

6. The template matching apparatus according to claim 3,

The input image signal, the partial image signal, and the template signal each include image information, The extracting means sequentially extracts the partial image signals while moving the extraction position along a raster scanning direction including a plurality of scanning lines, and also shifts the extraction position from a current scanning line to a next scanning line. At the time of moving, the moving amount from the current scanning line to the next scanning line is calculated in consideration of the similarity index at each position of the current scanning line soil, and based on the moving amount. A template matching apparatus, wherein the extraction position is moved.

7. The template matching apparatus according to claim 4,

 The input image signal, the partial image signal, and the template signal each include image information.

 The extracting means sequentially extracts the partial image signals while moving the extraction position along a raster scanning direction including a plurality of scanning lines, and also shifts the extraction position from a current scanning line to a next scanning line. At the time of moving, the amount of movement from the current scanning line to the next scanning line is calculated in consideration of the similarity index at each position on the current scanning line, and based on the moving amount A template matching apparatus, wherein the extraction position is moved.

8. Extraction means for sequentially extracting partial image signals for calculation from a plurality of different positions in the input image signal,

 First calculating means for calculating the similarity between the partial image signal extracted by the extracting means and a known template signal;

By comparing the similarities obtained by the first calculating means, the specifying means for specifying a matching position with the template signal in the input image signal; In order to determine the similarity between the partial image signal and a known template signal for each of the partial image signals, a second matching operation that performs a simple matching operation at a higher operation speed than the operation performed by the first arithmetic unit is performed. Arithmetic means; Determining means for determining whether or not to calculate the similarity between the partial image signal and a known template signal by the first calculating means based on a calculation result by the second calculating means. A template matching device, comprising:

9. The template matching apparatus according to claim 8, wherein:-each time the second image processing unit extracts the partial image signal by the extracting unit, a similarity between the partial image signal and a known template signal is generated. A template matching device for determining an index regarding gender.

10. The template matching apparatus according to claim 8,

 The second calculation means obtains a value related to similarity by comparing the skewness of the partial image signal and the skewness of the template signal,

 The determining means determines whether or not to perform the calculation in the first calculating means based on the value regarding the similarity.

 A template matching apparatus characterized in that:

11. The template matching apparatus according to claim 8,

 The second calculating means obtains a value related to similarity by comparing the average value of the intensity of the partial image signal and the average value of the intensity of the template signal, and the determination means determines the value based on the value related to the similarity. To determine whether or not to perform the calculation by the first calculating means.

 A template matching apparatus characterized in that:

1 2. Extraction means for sequentially changing the extraction position from the input image signal to extract a partial image signal for operation,

The partial image signal extracted by the extraction unit and a known template signal First calculating means for calculating the degree of similarity with

 Specifying means for specifying a matching position of the input image signal with the template signal by comparing the degree of similarity obtained by the first calculating means; and extracting the partial image signal by the extracting means Each time the partial image signal and a known template signal are indexed, a second calculating means for obtaining an index regarding the similarity between the partial image signal and the known template signal; In consideration of the movement amount determination means for determining the movement amount in the extraction means from the current extraction position to the next extraction position,

 Determining means for determining whether or not to perform a similarity calculation between the partial image signal and a known template signal by the first calculating means based on a calculation result by the second calculating means. A template matching apparatus characterized in that: -