JP4480980B2 - Image processing apparatus, image processing method, program, and recording medium - Google Patents

Description

  The present invention relates to an image processing technique for searching a region similar to a specific minute image (template) from search target images, and particularly, even when a plurality of template candidates are found in a search target image, The present invention relates to an image processing technique that enables a search with high accuracy.

As a method of searching for a template in the image, the degree of matching between the image area and the template at each position is calculated while shifting the template in the image, and is recorded in a table. A method for searching for a position is known (for example, see Non-Patent Document 1). In addition, as a conventional recognition processing technique, there is one disclosed in Non-Patent Document 2 shown below.
Makoto Nagao, “Digital Image Processing”, pp 306-310, Modern Science, 1995 Tetsuya Tsuji, Akira Suzuki, Akio Shio, Sakuichi Otsuka, Tomohiko Arikawa, “Sequential Batch Input Type Old Document Input System Based on Similar Pattern Search”, IEICE Technical Report, The Institute of Electronics, Information and Communication Engineers, VOL. 102, NO. 531, PRMU2002, pp129-145

  In the prior art, when there are a plurality of images having a high degree of similarity with a template, many candidate positions are detected, which increases the amount of processing and increases the time required for processing.

  The present invention has been made in view of such circumstances, and even when there are a plurality of search target images having high similarity to the template, the candidate positions can be narrowed down with high accuracy, and further the candidate positions can be narrowed down. Accordingly, an object of the present invention is to provide an image processing apparatus, an image processing method, a program, and a recording medium that can reduce the processing amount and perform high-speed recognition.

In the present invention, an area having a high degree of matching with a template in an image is detected with high accuracy and efficiency.
(1) Record a region having a high degree of matching with the template included in the search target image,
(2) Extract an arbitrary position from the recorded region with a high degree of matching and the region matching degree around it,
(3) By setting the position with the highest degree of matching in the extracted area as a candidate position where the template exists,
Even when there are a plurality of candidate positions where the template exists in the search target image, the candidate positions can be narrowed down to reliable candidate positions.

  Further, in the process of (2), before extracting the matching degree of the peripheral area, the area having a high matching degree is labeled by threshold processing and labeling, and the area having a high matching degree in the same label is used as a candidate position. By narrowing down, it becomes possible to increase the narrowing efficiency of candidate areas.

  The configuration of the present invention to be described below specifically shows how to calculate for narrowing down candidate regions in order to realize the basic configuration of the present invention.

In order to achieve this object, the image processing apparatus of the present invention provides:
A matching degree calculating means capable of calculating a matching degree with a template included in the search target image;
A matching degree recording means capable of recording the calculated matching degree in position coordinates corresponding to each pixel of the target image;
A candidate area narrowing means is provided that can calculate, as a candidate position, a position coordinate having the highest degree of matching in a peripheral area of an arbitrary position coordinate with respect to the recorded matching degree, and record the candidate position. Configure as follows.

  Further, in addition to this configuration, a feature extraction unit capable of performing feature extraction processing based on the image features of the search target image and the template when calculating the degree of matching between the search target image and the template is provided. The matching degree between the search target image and the template may be calculated by the matching degree calculating means using the feature extracted by the means.

  Further, in the candidate area narrowing-down means, the matching degree recorded in the matching degree recording means is threshold-processed with an arbitrary threshold as shown in FIG. 2, and the threshold processing result is labeled as shown in FIG. As shown in FIG. 4, labeling means may be provided for recording the position coordinates having the highest degree of matching within each label number of the labeling result as candidate positions.

  Furthermore, the candidate area narrowing means divides the matching degree recorded by the matching degree recording means or the candidate positions narrowed down by the labeling means at arbitrary intervals as shown in FIG. In this way, candidate area dividing means for recording the position coordinates having the highest degree of matching in the divided areas as candidate positions may be provided.

  At this time, as shown in FIG. 7, it is assumed that the division interval of the candidate area can be arbitrarily selected.

  Further, the candidate area dividing means may further include a divided area shifting means capable of providing redundancy to the divided areas by spatially moving the division intervals as shown in FIG.

  Further, in the candidate area narrowing means, arbitrary position coordinates are selected as shown in FIG. 9, and the position coordinates having the highest degree of matching in the position coordinates and the surrounding area are recorded as candidate positions as shown in FIG. Candidate area extracting means may be provided.

  At this time, the size of the peripheral region can be arbitrarily selected as shown in FIG.

In order to realize the above configuration, the image processing apparatus according to claim 1 is an image processing apparatus that searches and outputs a region similar to a template from search target images, and inputs the search target image and the template, A matching degree calculation means for calculating the degree of matching between the area in the search target image and the template based on the image features of the search target image and the template, and an area having a high degree of matching as a candidate area. After performing threshold processing based on the threshold on the matching degree recording means to be recorded and the candidate area recorded on the matching degree recording means, this processing result is labeled, and the matching degree within the same label number There performs labeling processing for recording the highest region as a candidate region in the label number, and the peripheral region of the candidate region of the labeling numbers Calculating a matching degree between PLATES, characterized in that it comprises, a candidate area narrowing means narrows down the matching degree higher region from among said candidate region and the peripheral region.

In the image processing apparatus according to the first aspect, the degree of matching with the template is calculated for the peripheral region of the candidate region, and the region having a high degree of matching is selected from the candidate region and the peripheral region. As a result, the processing speed increases. Moreover, a candidate area | region can be calculated for every label number, and a narrowing-down speed can be increased using this .

An image processing apparatus according to claim 2 is an image processing apparatus that searches and outputs a region similar to a template from search target images, and inputs the search target image and the template, and the search target image and the template are input. A matching degree calculation means for calculating a matching degree between the region in the search target image and the template based on the image feature and a matching degree record for recording a matching degree of the candidate region with a region having a high matching degree as a candidate region. And a candidate area dividing process of dividing the candidate area recorded in the matching degree recording means at regular intervals and recording the area having the highest matching degree among the divided areas as a candidate area in the divided area. Then, the degree of matching between the template area and the peripheral area of the candidate area in the divided area is calculated and adjusted from the candidate area and the peripheral area. Degree, characterized in that it comprises, a candidate area narrowing means narrows down the high region. The image processing apparatus according to claim 3 is an image processing apparatus for searching for and outputting a region similar to a template from search target images, wherein the search target image and the template are input, and the search target image and the template are input. A matching degree calculation means for calculating a matching degree between the region in the search target image and the template based on the image feature and a matching degree record for recording a matching degree of the candidate region with a region having a high matching degree as a candidate region. And the threshold value processing based on the threshold value for the candidate area recorded in the matching degree recording means, and then labeling the processing result, the area having the highest matching degree in the same label number A labeling process for recording as a candidate area for the label number is performed, and the candidate area recorded by the labeling process is divided at regular intervals. A candidate area dividing process is performed to record a region having the highest degree of matching among the divided areas as a candidate area in the divided area, and a degree of matching between the peripheral area of the candidate area and the template in the divided area is calculated, And candidate region narrowing means for narrowing down the candidate region and the peripheral region to regions having a high degree of matching.

The image processing apparatus according to claim 2 can calculate a candidate area for each of the divided areas, calculates a degree of matching with the template for a peripheral area of the candidate area, and a degree of matching from the candidate area and the peripheral area Since the narrowing is performed in a high area, the reliability of the narrowing becomes high, and the processing speed increases. The image processing apparatus according to claim 3, by dividing the area recorded by the labeling process, for calculating the candidate area for each divided area can be further increased narrowing rate.

According to a fourth aspect of the present invention, in the image processing apparatus according to the second or third aspect , the candidate area dividing unit can change the interval of the areas to be divided and can calculate a narrowing result for a plurality of division intervals. It is characterized by.

In the image processing apparatus according to the fourth aspect , it is possible to provide redundancy for narrowing down.

According to a fifth aspect of the present invention, in the image processing apparatus according to any one of the second to fourth aspects, the candidate area dividing unit may change the area to be divided by translating a reference position of the area to be divided. It is possible to calculate different narrowing results for a plurality of areas.

In the image processing apparatus according to the fifth aspect , it is possible to provide redundancy for narrowing down.

According to a sixth aspect of the present invention, in the first to fifth aspects, the matching degree calculating unit includes a feature extracting unit that creates a feature for identification from the image features of the input search target image and the template. The degree of matching between the region in the search target image and the template is calculated based on the created identification feature. Further, the candidate area narrowing means according to claims 1 to 5 performs extraction of an area while shifting an area of an arbitrary size from the candidate area, and narrows down candidate areas for narrowing down to the highest degree of matching from the extracted areas. Extraction means may be provided.

In the image processing device according to the sixth aspect, for example, the characteristics obtained from the intensity of the image based on the image characteristics and the direction of the intensity gradient and / or the direction of the intensity of the intensity and the gradient are used as the characteristics for identification. Can do. This is particularly effective when it is difficult to calculate the degree of matching from the image features of the input search target image and the template or when there is no difference in the degree of calculation. Further, by providing the candidate area narrowing means with the candidate area extracting means, it is possible to perform narrowing with higher reliability and increase the speed of narrowing down.

An image processing method according to claim 7 is an image processing method for searching and outputting a region similar to a template from search target images, wherein the search target image and the template are input, and the search target image and the template A matching degree calculation process for calculating the degree of matching between the region in the search target image and the template based on the image characteristics, and a matching degree recording process for recording the degree of matching of the candidate area with a region having a high matching degree as a candidate area; Then, after performing threshold processing based on the threshold for the candidate area recorded in the matching degree recording process, the processing result is labeled, and the area having the highest matching degree among the same label numbers is labeled. It performs labeling processing for recording the candidate area in the number, consistency degree between the peripheral region and the template of the candidate region of the labeling numbers Calculates, characterized in that and a candidate area narrowing process to narrow down the matching degree higher region from among said candidate region and the peripheral region.

In the image processing method according to the seventh aspect, since the degree of matching with the template is calculated for the peripheral area of the candidate area, and the candidate area and the peripheral area are narrowed down to areas with a high degree of matching, so the reliability of the narrowing is high. As a result, the processing speed increases. Moreover, a candidate area | region can be calculated for every label number, and a narrowing-down speed can be increased using this.

An image processing method according to claim 8 is an image processing method for searching and outputting a region similar to a template from search target images, wherein the search target image and the template are input, and the search target image and the template A matching degree calculation process for calculating the degree of matching between the region in the search target image and the template based on the image characteristics, and a matching degree recording process for recording the degree of matching of the candidate area with a region having a high matching degree as a candidate area; , Dividing the candidate area recorded in the matching degree recording process at regular intervals, and performing candidate area dividing processing for recording the area having the highest matching degree among the divided areas as a candidate area in the divided area , The degree of matching between the candidate area and the surrounding area of the candidate area in the divided area is calculated, and the degree of matching is calculated from the candidate area and the surrounding area. Characterized in that and a candidate area narrowing process to narrow down the high region. The image processing method according to claim 9 is an image processing method for searching and outputting a region similar to a template from search target images, wherein the search target image and the template are input, and the search target image and the template A matching degree calculation process for calculating the degree of matching between the region in the search target image and the template based on the image characteristics, and a matching degree recording process for recording the degree of matching of the candidate area with a region having a high matching degree as a candidate area; Then, after performing threshold processing based on the threshold for the candidate area recorded in the matching degree recording process, the processing result is labeled, and the area having the highest matching degree among the same label numbers is labeled. Perform a labeling process to record as a candidate area in the number, divide the candidate area recorded by the labeling process at regular intervals, Performing candidate region division processing for recording a region having the highest matching degree among the regions as a candidate region in the divided region, calculating a matching degree between a peripheral region of the candidate region and the template in the divided region, and And a candidate area narrowing-down process for narrowing down the area having a high degree of matching from the area and the peripheral area.

In the image processing method according to claim 8 , a candidate area can be calculated for each divided area, a matching degree with a template is calculated for a peripheral area of the candidate area, and a matching degree is determined from the candidate area and the peripheral area. Since the narrowing down is performed in the high region, the reliability of the narrowing becomes high and the processing speed increases. In the image processing method according to the ninth aspect, the area recorded by the labeling process is divided and a candidate area is calculated for each divided area, so that the narrowing-down speed can be further increased.

The image processing method according to a tenth aspect is the image processing method according to the eighth or ninth aspect , wherein the candidate area dividing step can change an interval between the divided areas, and can calculate a narrowing result for a plurality of dividing intervals. It is characterized by.

In the image processing method according to the tenth aspect , it is possible to provide redundancy for narrowing down.

An image processing method according to an eleventh aspect is the image processing method according to any one of the eighth to tenth aspects, wherein the candidate region dividing step changes a region to be divided by translating a reference position of the region to be divided. It is possible to calculate different narrowing results for a plurality of areas.

In the image processing method according to the eleventh aspect , it is possible to provide redundancy for narrowing down.

An image processing method according to a twelfth aspect of the present invention is the image processing method according to any one of the seventh to eleventh aspects, wherein the matching degree calculation step includes a feature extraction step of creating a feature for identification from the image feature of the input search target image and the template. The degree of matching between the region in the search target image and the template is calculated based on the created identification feature. In the candidate area narrowing-down process according to claims 7 to 11, a candidate area is extracted by shifting an area of an arbitrary size from the candidate area and narrowing down to the highest degree of matching from the extracted areas. An extraction process may be provided.

In the image processing method according to the twelfth aspect, for example, the intensity obtained from the intensity of the image based on the image characteristics and the direction of the intensity gradient and / or the characteristic obtained from the intensity of the intensity or the direction of the gradient is used as the identification characteristic. Can do. This is particularly effective when it is difficult to calculate the degree of matching from the image features of the input search target image and the template or when there is no difference in the degree of calculation. Further, since the candidate area narrowing process includes the candidate area extracting process, it is possible to perform narrowing with higher reliability and increase the speed of narrowing down.

A program according to a thirteenth aspect causes a computer to execute the image processing method according to any one of the seventh to twelfth aspects .

A computer-readable recording medium according to a fourteenth aspect is characterized in that the program according to the thirteenth aspect is recorded .

  With the configuration as described above, candidate positions in identification can be narrowed down, so that the processing speed can be increased by reducing the processing amount, and further, the reduction in the identification rate due to erroneous extraction can be prevented.

  As described above, according to the present invention, even when a plurality of candidate positions are detected when searching for an area similar to a specific template from the search target image, the candidates are highly accurate. The position can be narrowed down.

  Further, according to the present invention, by narrowing down candidate positions, the subsequent processing amount can be reduced, and high-speed search can be executed.

  Further, by narrowing down candidate positions, erroneous extraction in identification can be reduced, and identification accuracy can be improved.

  Hereinafter, the present invention will be described in detail according to embodiments.

  FIG. 1 illustrates an embodiment of an apparatus configuration of an image processing apparatus according to the present invention. In the figure, 101 is a matching degree calculating means, 102 is a matching degree recording means, 103 is a feature extracting means, 104 is a candidate area narrowing means, 105 is a labeling means, 106 is a candidate area dividing means, 107 is a divided area shifting means, 108 Is candidate area extraction means.

  Next, the detailed operation of each of these processing means will be described with reference to the flowcharts of FIGS.

  First, an embodiment in the case where one type of template is used will be described.

  In this embodiment, first, the matching degree calculation unit 101 operates. The matching degree calculation unit 101 reads the search target image and the template in ST1.

  Next, in ST2, when the identification is performed using the feature of the image itself, the process proceeds to ST4, and when the identification feature is extracted from the image feature, the process proceeds to ST3. At this time, taking into account the ratio of the size of the template to the search target image or the case where the template in the search target image is deformed, a plurality of search target images taking into account the deformation are created, and the plurality of created searches The processing of ST3 and ST4 can be performed on the target image.

  In ST3, the feature extraction unit 103 operates to convert the image feature into a feature for identification, for example, a feature such as a gradient strength calculated from the density of the image, and the process proceeds to ST4 using the converted feature. .

  In ST4, the degree of matching between the search target image and the template is calculated.

  For the calculation of the degree of matching, normalized cross-correlation (Masao Nagao, “Digital Image Processing” pp 306-310, Modern Science Co., Ltd., 1995), the characteristics of the search target image calculated by the feature extraction means 103 and the template By a distance value in the feature space of the image, a projection distance to a subspace approximated by a small number of feature vectors in the feature space (Hideyuki Tamura, “Computer Image Processing” pp 282-286, Ohm Publishing Co., 2002), and a kernel function Projection distance to the subspace created using the defined non-linear transformation (Eisaku Maeda, Hiroshi Murase, “Pattern Recognition by Kernel Nonlinear Subspace Method”, IEICE Poetry D-II, VOL.J82- D-II, NO.4, pp600-6122.1999), feature vector spaces such as haze estimation and maximum likelihood estimation That the occurrence probability (Richard O.Duda, Peter E.Hart, David G.Stock, "pattern identification", 84-101, new technology Communications, 2001) can be used.

  At this time, when a plurality of search target images are created in ST1, the degree of matching with the template is calculated for the plurality of search target images, and the plurality of search corresponding to the position coordinates of the search target image before deformation is calculated. The degree of matching between the target image and the template is compared, and the degree of matching that maximizes the degree of matching can be set as the degree of matching with the search target image at the position coordinates.

  Next, in ST5, the matching degree between the search target image and the template calculated by the matching degree calculation means 101 is recorded by the matching degree recording means 102, and the recorded matching degree is sent to the candidate area narrowing means 104.

  In ST6, the candidate area narrowing means 104 determines whether or not to perform labeling of the matching degree in order to narrow down the candidate area based on the matching degree between the search target image and the template.

  If the labeling process is performed for the degree of matching between the search target image and the template, the process proceeds to ST9. If the labeling is not performed, the process proceeds to ST7.

  In ST9, the labeling means 105 sets a threshold value for the degree of matching between the search target image and the template before labeling. After setting the threshold, threshold processing is performed in ST10 using the threshold, and labeling is performed in ST11.

  Next, in ST12, for each label of the labeling result, the position coordinate having the highest matching degree within the same label is recorded as a candidate position, and regions other than the candidate position within the same label are excluded from the candidates. Return to processing. At this time, when the same process is performed while changing the threshold, the process proceeds to ST9 again and the labeling is repeated.

  Next, in ST7, in order to narrow down the candidate area from the degree of matching between the search target image and the template recorded in the candidate area narrowing means 104, it is determined whether the matching degree is divided and processed. Here, if the search target image is processed by dividing the degree of matching with the template into regions, the process proceeds to ST13. If not, the process proceeds to ST8.

  In ST13, the candidate area dividing unit 108 designates a division interval for dividing the degree of matching between the search target image and the template, and proceeds to the process of ST14.

  In ST14, position coordinates for starting area division are designated, and in subsequent ST15, search target images and templates recorded in the candidate area narrowing means 104 based on the division intervals and division start position coordinates designated in ST13 and ST14. And the degree of matching with.

  In ST16 to ST17, for each of the divided blocks having the matching degree between the search target image and the template divided in ST15, the position coordinates having the highest matching degree in each area are calculated, and the calculated position coordinates are used as candidate areas. This is sent to the narrowing means 104, the position coordinates are recorded, areas other than the candidate positions in the same block are excluded from the candidates, and the process returns to ST7.

  In ST8, in order to narrow down the candidate area based on the degree of matching between the search target image and the template, the peripheral area including the position coordinates recorded in the candidate area narrowing means 104 is determined based on any position coordinates based on the degree of matching. Determine whether to extract and process.

  Here, when processing is performed by extracting the surrounding area including the position coordinates, the process proceeds to ST18. When the process is not performed, the position coordinates of the candidate area recorded in the candidate area narrowing means 104 are output as a result.

  In ST18, the candidate area extraction means 108 designates a range to be extracted as a peripheral area, and the process proceeds to ST19.

  In ST19, an arbitrary position coordinate is designated, and in ST20, the search target image and the template recorded in the candidate area narrowing means 104 are recorded based on the range as the peripheral area designated in ST18 and ST19 and the arbitrary position coordinate. A region to be extracted is selected from the matching degree.

  In ST21, based on the degree of matching between the extracted search target image and the template, the position coordinates having the highest degree of matching are sent to the candidate area narrowing means 104, the position coordinates are recorded, and the positions other than the candidate positions in the same area are recorded. The area is removed from the candidates, and it is determined whether or not all positions have been selected from the degree of matching between the search target image and the template recorded in the candidate area narrowing means 104 in ST22.

  If all positions have been selected, the process proceeds to ST8. Otherwise, the process proceeds to ST19.

  Next, an operation when there are a plurality of objects to be identified in the target and a plurality of templates corresponding to the plurality of objects are used will be described. Here, the plurality of objects are different figures to be identified existing in the target image such as “triangle”, “square”, and “circle” in the example of FIG. 16, and “a” in the example of FIG. , “I”, “U”, and the like, indicate different character types to be identified existing in the target image.

  In this embodiment, first, the matching degree calculation unit 101 operates. The matching degree calculation means 101 reads the image to be searched and a plurality of templates in ST1, and then extracts identification features from the image features to ST4 when performing identification using features of the image itself in ST2. If so, it is determined that the process proceeds to ST3.

  At this time, taking into account the ratio of the size of the template to the search target image, or the case where the template in the search target image is deformed, a plurality of search target images taking into account the deformation are created, and the created plurality of searches The processing of ST3 and ST4 can be performed on the target image.

  In ST3, the feature extraction unit 103 operates to convert the image feature into a distinguishing feature, for example, a feature such as a gradient strength calculated from the density of the image, and the process proceeds to ST4 using the converted feature.

  In ST4, the degree of matching between the search target image and the plurality of templates is calculated. Here, for the calculation of the matching degree, the same technique as in the embodiment in the case of using one type of template is used.

  Then, the matching level recording unit 102 records the matching level with the templates up to an arbitrary order in the descending order of matching level, and information indicating the template, and sends the information to the candidate area narrowing unit 104.

  In ST6, the candidate area narrowing means 104 first determines whether or not to perform labeling of the matching degree in order to narrow down the candidate area from the matching degree between the search target image and the plurality of templates.

  If the labeling process is performed for the degree of matching between the search target image and the plurality of templates, the process proceeds to ST9. If the labeling is not performed, the process proceeds to ST7.

  In ST9, before the labeling is performed, the labeling unit 105 sets a threshold value for the degree of matching between the search target image and the template. After setting the threshold, the threshold is used for the threshold processing in ST10, and the labeling is performed in ST11.

  Next, in ST12, for each label of the labeling result, the position coordinates having the highest matching degree within the same label are recorded as candidate positions in the candidate area narrowing means 104, and areas other than the candidate areas within the same label are recorded. It removes from a candidate area | region and returns to the process of ST6.

  Here, when calculating the position coordinates of the matching degree up to an arbitrary rank and the information indicating the template in descending order of matching degree, if the same template is listed as a candidate by the rank, the matching degree is higher. Position coordinates are candidates, and information on position coordinates with a low degree of matching is excluded from the ranking targets.

  At this time, when the same process is performed while changing the threshold, the process proceeds to ST9 again and the labeling is repeated.

  Next, in ST7, in order to narrow down the candidate area from the degree of matching between the search target image recorded in the candidate area narrowing means 104 and the plurality of templates, it is determined whether or not to process the matching degree.

  Here, if the search target surface image is processed by dividing the degree of matching with a plurality of templates, the process proceeds to ST13. Otherwise, the process proceeds to ST8.

  In ST13, the candidate area dividing unit 106 designates a division interval for dividing the degree of matching between the search target image and the plurality of templates, and proceeds to the process of ST14.

  In ST14, position coordinates for starting division are designated, and in subsequent ST15, search target images and a plurality of templates recorded in the candidate area narrowing means 104 based on the division interval and division start position coordinates designated in ST13 and ST14. And the degree of matching with.

  In ST16 to ST17, the position coordinates having the highest degree of matching are calculated in each region for each divided block of the degree of matching between the search target image and the plurality of templates divided in ST15, and the calculated position coordinates are calculated. This is sent to the candidate area narrowing means 104, the position coordinates are recorded, areas other than the candidate positions in the block are removed from the candidate areas, and the process returns to ST7.

  Here, when calculating the position coordinates of the matching degree up to an arbitrary rank and the information indicating the template in descending order of matching degree, if the same template is listed as a candidate by the rank, the matching degree is higher. Position coordinates are candidates, and information on position coordinates with a low degree of matching is excluded from the ranking targets.

  In ST8, in order to narrow down the candidate area from the matching degree between the search target image and the plurality of templates in the candidate area narrowing means 104, an arbitrary position coordinate is used as a reference from the matching degree recorded in the candidate area narrowing means 104. It is determined whether the peripheral area including the position coordinates is extracted and processed.

  Here, when processing is performed by extracting the surrounding area including the position coordinates, the process proceeds to ST18. When the process is not performed, the position coordinates of the candidate area recorded in the candidate area narrowing means 104 are output as a result.

  In ST18, the candidate area extraction means 108 designates a range to be extracted as a peripheral area, and the process proceeds to ST19.

  In ST19, an arbitrary position coordinate is specified, and in ST20, the search target image and a plurality of templates recorded in the candidate area narrowing means 104 based on the range as the peripheral area specified in ST18 and ST19, and the arbitrary position coordinate, A region to be extracted is selected from the degree of matching.

  In ST21, the position coordinates having the highest degree of matching are sent to the candidate area narrowing means 104 from the degree of matching between the extracted search target image and the plurality of templates, the position coordinates are recorded, and the candidate position of the same area is recorded. Remove areas other than.

  Here, when calculating the position coordinates of the matching degree up to an arbitrary rank and the information indicating the template in descending order of matching degree, if the same template is listed as a candidate by the rank, the matching degree is higher. Position coordinates are candidates, and information on position coordinates with a low degree of matching is excluded from the ranking targets.

  Subsequently, in ST22, it is determined whether or not all positions have been selected from the degree of matching between the search target image recorded in the candidate area narrowing means 104 and the plurality of templates.

  If all positions have been selected, the process proceeds to ST8. Otherwise, the process proceeds to ST19.

  In the present invention, some or all of the processing functions of the apparatus shown in FIG. 1 are configured as a program and realized using a computer, or the processing procedures shown in FIGS. 12 to 15 are configured as a program. It can be executed by a computer. In addition, a computer-readable recording medium such as a flexible disk, MO, ROM, or memory card can be used to store a program for realizing the processing functions of each unit by the computer or a program for causing the computer to execute the processing procedure. It can be recorded on a CD, DVD, removable disk, etc., stored, provided, and distributed via a communication network such as the Internet.

1 is a diagram illustrating a configuration example of an image processing apparatus according to the present invention. The figure which shows the example of a threshold value process. The figure which shows the example of a labeling process. The figure which shows the example which detects a candidate position from a labeling result. The figure which shows the example of candidate area | region division | segmentation. The figure which shows the example which narrows down a candidate area | region from the result of candidate area | region division | segmentation. The figure which shows the example of a change of a division | segmentation interval. The figure which shows the example of a change of a division | segmentation reference position. The figure which shows the example which extracts an area | region on the basis of arbitrary position coordinates. The figure which shows the example which changes the magnitude | size of an area | region. The figure which shows the candidate area | region narrowing down at the time of extracting arbitrary areas. The process flowchart of this invention. The process flowchart of this invention. The process flowchart of this invention. The process flowchart of this invention. The figure which shows the example of the object made into search object by this invention. The figure which shows the example of the character type made into search object by this invention.

Explanation of symbols

DESCRIPTION OF SYMBOLS 101 ... Matching degree calculation means 102 ... Matching degree recording means 103 ... Feature extraction means 104 ... Candidate area narrowing means 105 ... Labeling means 106 ... Candidate area dividing means 107 ... Divided area shifting means 108 ... Candidate area extracting means

Claims (14)

An image processing apparatus that searches and outputs a region similar to a template from search target images,
A matching degree calculating means for inputting a search target image and a template, and calculating a matching degree between a region in the search target image and the template based on an image feature of the search target image and the template;
An area with a high degree of matching is set as a candidate area, and a degree-of-match recording means for recording the degree of matching of this candidate area;
After the threshold value processing based on the threshold value is performed on the candidate area recorded in the matching degree recording unit, the processing result is labeled, and the area having the highest matching degree among the same label numbers is labeled with the label number. performs labeling processing for and recorded as the candidate regions in, calculates the matching degree between the peripheral region and the template of the candidate region of the labeling number, the narrowing in the region is higher matching degree among said candidate region and the peripheral region An image processing apparatus comprising: candidate area narrowing means for performing . An image processing apparatus that searches and outputs a region similar to a template from search target images,
A matching degree calculating means for inputting a search target image and a template, and calculating a matching degree between a region in the search target image and the template based on an image feature of the search target image and the template;
An area with a high degree of matching is set as a candidate area, and a degree-of-match recording means for recording the degree of matching of this candidate area;
The candidate area recorded in the matching degree recording means is divided at regular intervals, and a candidate area dividing process is performed in which the area having the highest degree of matching among the divided areas is recorded as a candidate area in the divided area. And a candidate area narrowing means for calculating a degree of matching between a candidate area and a template in a peripheral area of the candidate area and narrowing down the candidate area and the peripheral area to a region having a high degree of matching. Image processing device. An image processing apparatus that searches and outputs a region similar to a template from search target images,
  A matching degree calculating means for inputting a search target image and a template, and calculating a matching degree between a region in the search target image and the template based on an image feature of the search target image and the template;
  An area with a high degree of matching is set as a candidate area, and a degree-of-match recording means for recording the degree of matching of this candidate area;
  After the threshold value processing based on the threshold value is performed on the candidate area recorded in the matching degree recording unit, the processing result is labeled, and the area having the highest matching degree among the same label numbers is labeled with the label number. The candidate area recorded by the labeling process is divided at regular intervals, and the area with the highest degree of matching is recorded as a candidate area in the divided area. Candidate area narrowing means that performs candidate area dividing processing, calculates the degree of matching between the peripheral area of the candidate area and the template in the divided area, and narrows down the candidate area and the peripheral area to areas having a high degree of matching An image processing apparatus comprising: The candidate area dividing means includes:
The image processing apparatus according to claim 2, wherein an interval between divided areas can be changed, and a narrowing result can be calculated for a plurality of divided intervals. The candidate area dividing means includes:
The reference position of the divided regions can be changed region to divide by moving parallel to any one of claims 2-4, characterized in that it is possible to calculate the different refined result of a plurality of regions The image processing apparatus described. The matching degree calculation means includes:
Comprising a feature extraction means for creating a feature for identification from the image features of the input search target image and the template,
6. The image processing apparatus according to claim 1 , wherein a degree of matching between a region in a search target image and a template is calculated based on the created identification feature. An image processing method for searching and outputting a region similar to a template from search target images,
A matching degree calculation process of inputting a search target image and a template, and calculating a matching degree between a region in the search target image and the template based on an image feature of the search target image and the template;
A matching degree recording process in which an area having a high degree of matching is set as a candidate area and the matching degree of the candidate area is recorded,
After the threshold processing based on the threshold is performed on the candidate region recorded in the matching degree recording process, the processing result is labeled, and the region having the highest matching degree among the same label numbers is labeled with the label number. performs labeling processing for and recorded as the candidate regions in, calculates the matching degree between the peripheral region and the template of the candidate region of the labeling number, the narrowing in the region is higher matching degree among said candidate region and the peripheral region And a candidate area narrowing-down process to be performed . An image processing method for searching and outputting a region similar to a template from search target images,
A matching degree calculation process of inputting a search target image and a template, and calculating a matching degree between a region in the search target image and the template based on an image feature of the search target image and the template;
A matching degree recording process in which an area having a high degree of matching is set as a candidate area and the matching degree of the candidate area is recorded,
The candidate area recorded in the matching degree recording process is divided at regular intervals, and a candidate area dividing process is performed in which the area having the highest degree of matching among the divided areas is recorded as a candidate area in the divided area. A candidate area narrowing process that calculates a degree of matching between a candidate area and a template in a peripheral area of the candidate area, and narrows down the candidate area and the peripheral area to a region having a high degree of matching. Image processing method. An image processing method for searching and outputting a region similar to a template from search target images,
A matching degree calculation process of inputting a search target image and a template, and calculating a matching degree between a region in the search target image and the template based on an image feature of the search target image and the template;
A matching degree recording process in which an area having a high degree of matching is set as a candidate area and the matching degree of the candidate area is recorded,
After the threshold processing based on the threshold is performed on the candidate region recorded in the matching degree recording process, the processing result is labeled, and the region having the highest matching degree among the same label numbers is labeled with the label number. The candidate area recorded by the labeling process is divided at regular intervals, and the area with the highest degree of matching is recorded as a candidate area in the divided area. Candidate area narrowing process that performs candidate area division processing, calculates the degree of matching between the peripheral area of the candidate area and the template in the divided area, and narrows down the candidate area and the peripheral area to areas having a high degree of matching And an image processing method . The candidate area dividing process includes:
The image processing method according to claim 8 or 9 , wherein the interval between the divided regions can be changed, and a narrowing result can be calculated for a plurality of divided intervals. The candidate area dividing process includes:
The reference position of the divided regions can be changed region to divide by moving parallel to one of claims 8 to 10, characterized in that it is possible to calculate the different refined result of a plurality of regions The image processing method as described. The matching degree calculation process includes:
It has a feature extraction process that creates features for identification from the image features of the input search target image and template,
The image processing method according to claim 7 , wherein the degree of matching between the region in the search target image and the template is calculated based on the created identification feature. An image processing program for causing a computer to execute the image processing method according to any one of claims 7 to 12 . 14. A computer-readable recording medium on which the program according to claim 13 is recorded .

Images