JP2009284084A - Image collating method, image collating apparatus, image data output apparatus, program, and storage medium - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To achieve an image collating apparatus which determines similarity to a non-original image which is formed by copying an original image with high precision when the original image is a color image. <P>SOLUTION: In an image collating apparatus 101, a document collating section 2 extracts shape features indicating the features about the shape of an image from input image data, and compares the shape features with those of a registered image. A color similarity determination section 4 determines similarity of an input image to the registered image, extracts color features indicating the features about the color of an image from the input image data, compares the color features with those of the registered image, and determines similarity of the input image to the registered image. Furthermore, a saturation determination section 43 obtains saturation information of the input image from the input image data. A general similarity determination section 5 determines similarity of the input image to the registered image, based on the determination results and the saturation information. <P>COPYRIGHT: (C)2010,JPO&INPIT

Description

  The present invention relates to an image matching method, an image matching device, an image data output processing device, a program, and a storage medium that process image data of a color image.

  2. Description of the Related Art Conventionally, a document image matching technique for determining the similarity between images is known, and this document image matching technique is used in various ways in image data processing.

  For the determination of the similarity between image data, for example, a keyword is extracted from an image using an OCR (Optical Character Reader), the similarity of the image is determined based on the extracted keyword, and the image for which the similarity is determined is a ruled line. A method for extracting ruled line features and determining image similarity, replacing character strings of image data with points, and obtaining the positional relationship of points (feature points) as feature quantities A method for determining the similarity of images has been proposed.

  For example, in Patent Document 1, a descriptor is generated from the characteristics of an input document, and the descriptor is recorded with the descriptor database indicating a list of documents including the descriptor and the generated characteristics. A technique for matching with a document in a database is disclosed. The descriptor is chosen to be invariant to distortions caused by digitization of the document and differences between the input document and the matching document in the document database.

  In this technology, when the descriptor database is scanned, votes for each document in the document database are accumulated, and one document with the maximum number of votes or a document whose number of votes exceeds a certain threshold is registered or similar. Is supposed to be extracted as

  In Patent Document 2, a plurality of feature points are extracted from a digital image, a set of local feature points is determined for each extracted feature point, and a subset of feature points is selected from each determined set Then, as an amount to characterize each selected subset, invariants for geometric transformation are obtained based on a plurality of combinations of feature points in the subset, and feature amounts are calculated by combining the obtained invariants. A technique for searching for a document image corresponding to the digital image by voting on a document image in a database having a calculated feature amount is disclosed.

  Japanese Patent Application Laid-Open No. 2004-228561 discloses a method for determining similarity by extracting image features from image color information. According to this, referring to the color group classification table that converts the RGB value of each pixel of the image into an HSL value, and classifies the colors into a plurality of color groups based on the values of the plurality of color elements specifying the color, The color of each pixel constituting the image is counted for each color group, the average color of each color group is obtained based on the value of the color element of the pixel counted in the color group, and the pixel counted in the color group is the image The occupancy ratio, which is the ratio of the entire image, is calculated for each color group, and the hue, saturation, and lightness representing the average color of the color group, and the occupancy ratio of the color group are extracted for each color group as the color characteristics of the image. .

  The similarity between the images is calculated using the color feature for each color group extracted in this way and the color feature of the comparison target image. The similarity is compared with one color feature and all the color features of the comparison image, the hue similarity, the saturation similarity and the lightness similarity in each color feature combination are calculated, and the similarity weight Is used to determine the degree of similarity for one color feature. Then, select the similarity with the maximum absolute value from the calculated similarities, multiply the maximum similarity obtained for each of the other color features by the occupation rate corresponding to this similarity, The sum of the color features is calculated as the similarity to the comparative control image.

  Also, in Patent Document 4, the feature amount (color feature amount, character feature amount) of the input image is compared with the feature amount of all pages of the original document to obtain the similarity, and the similarity is the first. It describes that document IDs and pages of electronic documents that are higher than the threshold are extracted as document candidates, and when there are a plurality of document candidates, display is performed to prompt the user to select. In addition, it is described that when the average similarity of the searched pages is lower than the first threshold value, it may be deleted from candidates and narrowed down.

Further, in Non-Patent Document 1, when extracting a writing area from a document, not only a monochrome document but also a color document, an image is subjected to color clustering processing in an RGB color space, and connected for each color obtained by this processing. It is described that collation determination is performed by calculating feature points of components.
Japanese Patent Laid-Open No. 7-282088 (released on October 27, 1995) International Publication No. 2006/092957 pamphlet (published on September 8, 2006) Japanese Patent Laid-Open No. 11-96364 (published on April 9, 1999) JP 2006-31181 A (published February 2, 2006) Tomohiro Nakai, Koichi Kise, Masakazu Iwamura: “Writing Extraction from Documents Using Registration Based on Localization of Feature Points” (Technical Research Report of IEICE “Pattern Recognition and Media Understanding (PRMU) ), March 2007 ")

  However, in the conventional technique described above, when the original image is a color image, the similarity between the registered image in which the original image is registered and the non-original image formed by copying the image is accurately determined. There is a problem that can not be done.

  That is, in the methods of Patent Documents 1 and 2, it is not clearly described which signal is used for the collation determination, but by using a binarized signal, a luminance signal or a G signal is used. Is presumed to be used. In such similarity determination, only the image component (feature related feature) is determined, and color information (color related feature) is not taken into consideration, so the image component is similar, but the color is completely different. Even different documents are determined to have high similarity.

  For example, in the case of an application form, a document image with a similar layout (similarity based on luminance information) is often printed with different color characters and ruled lines so that the difference in type can be easily identified. Is called. In the methods of Patent Literatures 1 and 2, it is determined that these different types of application forms have high similarity.

  On the other hand, as in Patent Document 3, when the similarity of the image is determined only by the color information, the similarity of the constituent elements of the image is not taken into consideration. There is a risk of being determined to be high. In addition, the original document is a color image, which may be copied in black and white or transmitted by facsimile to create a non-original document as a black and white image. If sex is judged, it will be judged that it is a completely different image.

  In addition, the method of Patent Document 3 has a problem in that it takes a very long time to perform collation determination because it is necessary to obtain similarity for each comparison target image (corresponding to a registered image) stored in the database. .

  On the other hand, in the methods of Patent Document 4 and Non-Patent Document 1, since both the component of the image and the color information are used for the determination, there is no problem as in Patent Documents 1 to 3 using only one of them. However, for example, for three sheets of a red ruled original, a green-formed original, and a black ruled original having the same degree of similarity in shape (character and ruled line shape), 1) a red ruled original It is determined that both the originals of green and green ruled lines and 2) originals of red ruled lines and originals of black ruled lines are low in color similarity, and cannot give priority to 1) and 2).

  The present invention has been made in view of the above problems, and when an original image is a color image, the similarity with a non-original image formed by copying the image is accurately determined. An object of the present invention is to provide an image collation method, an image collation apparatus, an image data output processing apparatus, a program, and a storage medium.

  The applicant of this application has many cases where a color manuscript is achromatized by monochrome copying or monochrome printing, compared to the case of color copying or color printing using other colors. If 1) the document with red ruled lines and the document with green ruled lines, and 2) the document with red ruled lines and the document with black ruled lines are judged to have a higher similarity in 2) than in 1) Focusing on the fact that is realistic, the present invention has been made.

  In order to solve the above-described problem, the image matching apparatus of the present invention extracts a shape feature amount indicating a feature related to the shape of an image from input image data, and compares it with the shape feature amount of a registered image registered. A shape similarity determination unit that determines the similarity between the input image and the registered image; and a color feature amount indicating a feature relating to the color of the image from the input image data; A color similarity determination unit that determines the similarity between the input image and the registered image, and by extracting the saturation from the input image data, whether the input image is chromatic In the saturation determination unit that determines whether the color is achromatic, the determination result of the shape similarity determination unit, the determination result of the color similarity determination unit, and the determination result of the saturation determination unit, the input image and Similarity total for determining similarity with the registered image It is characterized by comprising a determining unit.

  In the similarity determination based on the shape feature amount, for example, in the case of an image that uses different colors for characters and ruled lines in a similar layout such as an application form, the colors are different because the features related to the shape are similar. Also, it is determined that the similarity is high.

  On the other hand, in the similarity determination based on the color feature amount, it is determined that the degree of similarity is high when the colors are similar even if the images look quite different. In addition, the original is a color image, which may be copied in black and white or transmitted by facsimile to create a non-original image of the black and white image. Then, it will be determined that the similarity is low.

  Further, even when both the image component and the color information are used for the determination, in the configuration in which the saturation information is not used for the determination, the similarity regarding the shape (the shape of the characters and the ruled lines) is high. About three sheets of a red ruled original, a green-formed original, and a black ruled original, both of which are the same, 1) a red ruled original and a green ruled original 2) a red ruled original and black The ruled line document also has a low color similarity, and it was not possible to give priority to 1) and 2).

  On the other hand, according to the above configuration, the similarity between the input image and the registered image is determined based on the similarity determination result (shape feature amount similarity) based on the shape feature amount indicating the feature related to the shape of the image and the color of the image. The determination is made comprehensively by using the similarity determination result (similarity of the color feature amount) based on the color feature amount indicating the feature and the saturation information of the image.

  Therefore, even if the input image and the registered image are color images such as application forms that have similar characters and ruled line colors that could not be identified by similarity determination based on shape feature amounts, The similarity can be judged well.

  Similarly, the input image and the registered image are color images with similar colors that could not be identified by the similarity determination based on the color feature amount, or the registered image is a color image and the input image is black and white. Even for black and white images formed by copying or facsimile transmission, similarity can be determined with high accuracy.

  Then, by determining using saturation information in addition to the component and color information of the image, 2) has higher similarity than 1) to 1) and 2) as described above. This makes it possible to carry out judgments that are more practical.

  As a result, when the original image is a color image, the similarity with a non-original image formed by copying the original image can be accurately determined.

  Note that the image data of the document input to the image processing apparatus is, for example, input necessary information using a computer (software) in the format of image data obtained by reading the document with a scanner or electronic data. Electronic data created by That is, for example, an image printed or described on paper and digitized, and an image directly created as electronic data (such as an electronic application form).

  In the image collation device of the present invention, the color similarity determination unit further uses a hue as the color feature amount, and a hue extraction unit that extracts a hue from input image data and the hue extraction unit extract the hue. A hue similarity determination unit that compares the hue of the input image with the hue of the registered image to determine the similarity between the input image and the registered image may be provided.

  According to this, the color feature amount of the image can be accurately extracted by using the hue as the color feature amount. In addition, although there is also lightness (luminance) as a color feature amount, since the lightness is used for determination of similarity by shape feature amount, the same parameter can be obtained by using a parameter different from the determination of shape feature amount. The determination accuracy can be made higher than that used.

  In the image collation device according to the present invention, the similarity comprehensive determination unit further determines that the input image is achromatic and chromatic if the similarity based on the shape feature amount is similar and the similarity based on the color feature amount is low. Then, it can also be set as the structure which determines that the achromatic one has high similarity with the said registration image.

  When judging the similarity between the registered image and the input document comprehensively by combining the judgment based on the shape feature amount and the judgment based on the color feature amount using the hue, the similarity based on the shape feature amount is similar and the similarity based on the hue Assuming that there are two input images that have similar low values, it is determined that neither of these two input images is similar to the registered image.

  For example, for example, an image composed of green characters and ruled lines and a black character determined to be similar in the determination by the shape feature amount to an image composed of red characters and ruled lines. The two images of the ruled line image are determined not to be similar to the image of the red color character or ruled line in the similarity determination by hue.

  However, in practice, an original image of a color image is often copied as a monochrome image or transmitted by facsimile and used as an achromatic non-original image. In addition, as in the application form described above, there are many document images in which characters of different colors and ruled lines are used for document images having a similar layout (high similarity due to shape feature amounts). Furthermore, it is almost unthinkable to convert a color document from its original color to another chromatic color.

  Therefore, in the previous example, it is better to determine that the similarity of a black text / ruled line document is higher than that of a green text / ruled line document in consideration of the usage status of the document image. Become.

  In the above configuration, when the similarity by the shape feature amount is the same and the similarity by the color feature amount using the hue is low, the achromatic image and the chromatic image are more similar to the achromatic image. Since the original color image is copied in black and white or sent by facsimile, the achromatic non-original image is formed. With respect to an image such as the above-mentioned application form in which is used, determination according to the user's feeling is possible, and determination with high accuracy is possible.

  In the image matching device of the present invention, the similarity total determination unit further includes a case where the similarity based on the shape feature amount is low and the similarity based on the color feature amount is high, and a case where the similarity based on the shape feature amount is high. In the case where the similarity due to is low, the case where the similarity based on the shape feature amount is high and the similarity based on the color feature amount is low may be determined as having a high similarity with the registered image.

  According to this, the similarity between the registered image and the input image can be determined more accurately by prioritizing the similarity determination result based on the shape feature amount between the similarity based on the shape feature amount and the similarity based on the color feature amount. can do.

  In the image matching device of the present invention, the color similarity determination unit refers to the determination result by the shape similarity determination unit when extracting the color feature amount, and the similarity by the shape feature amount is high. It is also possible to adopt a configuration in which color feature amounts are extracted only for registered images that are determined to be.

  According to this, even when there are many registered images that need to be determined to be similar to the input image, it is possible to narrow down the registered images from which the color feature amount is extracted. The sex determination can be completed promptly.

  In the image matching device of the present invention, the shape similarity determination unit further extracts a feature point calculation unit by extracting a connected portion of the input image from input image data, and the feature point calculation. A shape feature amount calculation unit that calculates a shape feature amount of the input image based on a relative position between the feature points calculated by the unit, a shape feature amount of the input image calculated by the shape feature amount calculation unit, and the registration It may be configured to include a shape similarity calculation unit that compares the shape feature amount of the image and calculates the similarity of the shape between the input image and the registered image.

  In order to solve the above-described problem, the image processing apparatus according to the present invention is configured to extract the similarity between the first image and the second image from the image data of the first image and the second image. A shape similarity determination unit that compares and determines a shape feature amount indicating a feature, and extracts the similarity between the first image and the second image from the image data of the first image and the second image A color similarity determination unit that compares and determines a color feature amount indicating a color-related characteristic of the obtained image, extracts saturation from each image data of the first image and the second image, A saturation determination unit that determines whether the second image is chromatic or achromatic, a determination result of the shape similarity determination unit, a determination result of the color similarity determination unit, and the saturation determination unit The similarity total determination unit that determines the similarity between the first image and the second image based on the determination result by It is characterized in that it comprises.

  According to this, the similarity between the first and second images is determined based on the similarity determination result based on the shape feature amount indicating the feature relating to the shape of the image, and the similarity based on the color feature amount indicating the feature relating to the color of the image. The determination result is comprehensively determined using the determination result and the saturation information of the image.

  Therefore, the first and second images are color images such as application forms that cannot be identified by the similarity determination based on the shape feature quantity and that have similar elements with different characters and ruled line colors. In addition, it is possible to accurately determine the similarity even between color images having similar colors or images obtained by copying the color image in black and white, which cannot be identified by the similarity determination based on the color feature amount. it can.

  Further, by determining using saturation information in addition to the component and color information of the image, 2) is higher than 1) in comparison with 1) and 2) as described above. This makes it possible to carry out judgments that are more practical.

  As a result, when the original image is a color image, the similarity with a non-original image formed by copying the original image can be accurately determined.

  In order to solve the above problems, an image data output processing device of the present invention is an image data output processing device that performs output processing on image data, and includes the above-described image collation device of the present invention. Yes.

  As already described as the image collating device, according to the above configuration, when the original image is a color image, the similarity with a non-original image formed by copying the image is accurately determined. Therefore, it is possible to provide an image data output processing device with high matching accuracy for color images by mounting such an image matching device.

  In order to solve the above problems, the image processing method of the present invention extracts a shape feature amount indicating a feature related to the shape of an image from input image data, and compares it with the shape feature amount of a registered image registered. , A shape similarity determination step for determining similarity between the input image and the registered image; and a color feature amount indicating a feature relating to the color of the image is extracted from the input image data; A color similarity determination step for determining the similarity between the input image and the registered image, and extracting the saturation from the input image data, so that the input image is chromatic In the saturation determination step for determining whether the color is achromatic, the determination result in the shape similarity determination step, the determination result in the color similarity determination step, and the determination result in the saturation determination step, the input image and Judge similarity with the registered image It is characterized by including a similarity comprehensive judgment process that.

  As already described as the image collating device, according to the above configuration, when the original image is a color image, the similarity with the non-original image formed by copying the original image can be accurately determined. This makes it possible to carry out judgments that are more practical.

  Further, the image collating apparatus may be realized by a computer. In this case, by causing the computer to operate as the respective units, a program for realizing the image collating apparatus by the computer, and a computer reading that records the program. Possible recording media are also included in the scope of the present invention.

  As described above, the image collating apparatus of the present invention extracts a shape feature amount indicating a feature related to the shape of an image from input image data, compares it with the shape feature amount of a registered image, and inputs the input A shape similarity determination unit for determining the similarity between the input image and the registered image; and extracting a color feature amount indicating a feature relating to the color of the image from the input image data; A color similarity determination unit that determines the similarity between the input image and the registered image, and extracts saturation from the input image data, so that the input image is chromatic or achromatic The input image and the registered image according to the determination result of the saturation determination unit that determines whether there is a determination result of the shape similarity determination unit, the determination result of the color similarity determination unit, and the determination result of the saturation determination unit A similarity total determination unit for determining similarity to Example was a configuration.

  Accordingly, when the original image is a color image, an image processing apparatus capable of accurately determining the similarity with a non-original image formed by copying the original image is provided, and collation with the color image is provided. There is an effect that a highly accurate image forming apparatus can be provided.

  Embodiments of the present invention will be described below with reference to the drawings. FIG. 1 is a block diagram showing a configuration of an image collating apparatus 101 in the present embodiment. The image collating apparatus 101 is provided in, for example, a digital color copying machine (image data output processing apparatus) 102 shown in FIG.

  The document (original) to be processed by the image matching device 101 is not particularly specified, but the image matching device 101 has a function of determining the similarity between images and can register images in advance. In this configuration, the similarity between the registered image and the image of the document input to be processed can be determined.

  Hereinafter, an image of a document that is registered in advance is referred to as a registered image, and an image of a document that is input for processing by the image collating apparatus 101 such as copying, facsimile, or filing is referred to as an input image.

  The image collation apparatus 101 determines the similarity between the registered image and the input image input to be processed, and outputs a determination signal indicating the determination result. When the image collation apparatus 101 of this embodiment determines that the increase in input image data is similar to the registered image, an output process (image forming process in a color copying machine) is performed on the image data of the input image according to the output determination signal. Output processing such as prohibition, image quality reduction, or image data filing is controlled.

  As shown in FIG. 1, the image matching device 101 includes a control unit 1, a document matching unit (shape similarity determination unit) 2, a color similarity determination unit 4, a similarity total determination unit 5, and a memory (storage unit) 3. It has.

  The document collation unit 2 extracts a shape feature amount indicating a feature related to the shape of the image from the input image data, compares it with the shape feature amount of the registered image, and compares the input input image with the input image data. The similarity with the registered image is determined. The determination result is output to the similarity comprehensive determination unit 5. In the present embodiment, a function for registering a registered image is also added to the document matching unit 2, and image data for registration is also input.

  The document collation unit 2 includes a feature point calculation unit 11, a feature amount calculation unit 12, a voting processing unit 13, a document similarity determination processing unit 14, and a registration processing unit 15.

  When image data of an input image or a registered image is input, the feature point calculation unit 11 extracts a connected portion of a character string or a ruled line from the input image data, and calculates the center of gravity of the connected portion as a feature point. In the present embodiment, the feature point calculation unit 11 also calculates the coordinates of each feature point.

  The feature amount calculation unit (shape feature amount calculation unit) 12 uses the feature points calculated by the feature point calculation unit 11 and is invariable with respect to rotation, enlargement, and reduction, that is, rotation and translation of the document image. Then, a feature amount (hash value) that is a parameter that is invariant to a geometric change including enlargement / reduction is calculated. In order to calculate the feature amount, a feature point near the target feature point is selected and used. This feature amount is a shape feature amount indicating a feature related to the shape of the image.

  The voting processing unit 13 is registered in a hash table described later using the hash value calculated by the feature amount calculation unit 12 for each feature point calculated by the feature point calculation unit 11 from the image data of the input image in the collation mode. Vote for registered images. The voting processing unit 13 votes for a registered image having the same hash value as the hash value of the image data of the input image.

  The document similarity determination processing unit 14 determines whether the input image is similar to the registered image based on the voting process result of the voting processing unit 13. The shape similarity calculation unit according to the present invention includes the voting processing unit 13 and the document similarity determination processing unit 14.

  The registration processing unit 15 registers an ID representing a registered image according to the hash value calculated by the feature amount calculation unit 12 for each feature point calculated by the feature point calculation unit 11 from the registered image data in the registration mode. is there.

  In the document collation unit 2, the voting processing unit 13 and the document similarity determination processing unit 14 perform processing in the collation mode, and do not perform processing in the registration mode. Conversely, the registration processing unit 15 performs processing in the registration mode and does not perform processing in the collation mode.

  The color similarity determination unit 4 extracts a color feature amount indicating a feature related to the color of the image from the input image data, compares it with the color feature amount of the registered image, and inputs the input image And the similarity between the registered image and the registered image. The determination result is output to the similarity comprehensive determination unit 5. In the present embodiment, the color similarity determination unit 4 has a function of registering a registered image, and image data for registration is also input. In the present embodiment, the hue is used as the color feature amount, but is not limited to the hue.

  The color similarity determination unit 4 includes a hue information extraction unit 41, a hue similarity determination processing unit 42, a saturation determination processing unit 43, and a hue information registration processing unit 44.

  The hue information extraction unit (hue extraction unit) 41 extracts hue information from image data. The hue similarity determination processing unit (hue similarity determination unit) 42 compares the hue information extracted by the hue information extraction unit 41 with the hue information of the registered image in the matching mode, and compares the input image with the registered image. Are similar to each other.

  The saturation determination processing unit (saturation determination unit) 43 calculates saturation information from the image data of the input image in the collation mode, and whether the input image is “chromatic” or “achromatic” based on the calculated saturation information. Is determined.

  The hue information registration processing unit 44 registers the hue information extracted by the hue information extraction unit 41 in association with an ID representing a registered image in the registration mode.

  In the color similarity determination unit 4, the hue similarity determination processing unit 42 performs processing in the collation mode and does not perform processing in the registration mode. Conversely, the hue information registration processing unit 44 performs processing in the registration mode and does not perform processing in the collation mode.

  The similarity comprehensive determination unit 5 determines the similarity (similarity) between the input image and the registered image based on the determination results of the document similarity determination processing unit 14 and the hue similarity determination processing unit 42 in the collation mode. Is. In the present embodiment, the determination result of the saturation determination processing unit 43 is also used.

  A control unit (CPU) 1 controls access to the above-described units and the memory 3 in the image collating apparatus 101. The memory 3 is a working memory when the above-described units in the image collating apparatus 101 perform processing, and an index (ID; for identifying an image to be registered in the registration processing). Etc.) where various information is registered.

  Hereinafter, each part of the image collating apparatus 101 will be described in detail with reference to the drawings. First, the document collation unit 2 in the image collation apparatus 101 will be described.

  As shown in FIG. 3, the feature point calculation unit 11 in the document matching unit 2 includes the achromatic processing unit 21, the resolution conversion unit 22, the MTF processing unit 23, the binarization processing unit 24, and the like. A centroid calculating unit 25 is provided. FIG. 3 is a block diagram showing a configuration of the feature point calculation unit 11.

  When the input image data, which is image data such as a registered image or an input image, is a color image, the achromatic processing unit 21 achromatizes the input image data and converts it into a brightness or luminance signal. For example, the luminance Y is obtained from the following formula (1).

The achromatization process is not limited to the above method, and the RGB signal is converted into a CIE 1976 L ^* a ^* b ^* signal (CIE: Commission International de l'Eclairage, L ^* : brightness index, a ^* , b ^* : color. (Degree index).

  When the input image data is optically scaled by the image input device, the resolution conversion unit 22 scales the input image data again so as to obtain a predetermined resolution. The image input device is, for example, a scanner that reads an image of a document and converts it into image data, and the color image input device 111 corresponds to this in the digital color copying machine shown in FIG.

  Further, the resolution conversion unit 22 is also used as a resolution conversion unit for lowering the resolution from the resolution read by the image input device with the same magnification setting in order to reduce the data processing amount in the subsequent stage. For example, image data read at 600 dpi (dot per inch) is converted to 300 dpi.

  The MTF processing unit 23 is used to absorb the influence caused by the difference in the spatial frequency characteristics of the image input device for each type of image input device. In other words, the image signal output from the CCD provided in the image input device includes optical components such as lenses and mirrors, aperture aperture of the light receiving surface of the CCD, transfer efficiency and afterimage, integration effects due to physical scanning, and uneven operation. This causes degradation of MTF. Due to the deterioration of the MTF, the read image is blurred. Therefore, the MTF processing unit 23 performs a process of repairing the blur caused by the degradation of the MTF by performing an appropriate filter process (enhancement process). Further, it is also used to suppress high-frequency components that are not necessary for processing in the feature point extraction unit 31 of the subsequent feature amount calculation unit 12. That is, the enhancement and smoothing processing is performed using the mixing filter. In addition, the filter coefficient of this mixing filter is shown in FIG. 4, for example.

  The binarization processing unit 24 binarizes the image data by comparing the brightness value (brightness signal) or brightness value (brightness signal) of the image data achromatized by the achromatization processing unit 21 with a threshold value. The binarized image data (registered image, binary image data of the input image) is stored in the memory 3.

  The center-of-gravity calculation unit 25 performs labeling (labeling processing) on each pixel of the image data binarized by the binarization processing unit 24 (for example, image data represented by “1” and “0”). Do. In this labeling, the same label is given to the pixels showing the same value among the two values. Next, a connected area, which is an area composed of a plurality of pixels formed by connecting pixels with the same label, is specified. Next, the center of gravity of the identified connected area is extracted as a feature point, and the extracted feature point is output to the feature amount calculation unit 12. Here, the feature point can be represented by coordinate values (x coordinate, y coordinate) in the binary image, and the coordinate value of the feature point is also calculated and output to the feature amount calculation unit 12.

  FIG. 5 is an explanatory diagram showing an example of a connected area extracted from binarized image data and the center of gravity of this connected area. The connected area corresponding to the letter “A” and the center of gravity of the connected area (features) Point). FIG. 6 is an explanatory diagram illustrating an example of each centroid (feature point) of a plurality of connected regions extracted from a character string included in binarized image data.

  As shown in FIG. 7, the feature amount calculation unit 12 includes a feature point extraction unit 31, an invariant calculation unit 32, and a hash value calculation unit 33. FIG. 7 is a block diagram showing a configuration of the feature amount calculation unit 12.

  When there are a plurality of feature points calculated by the feature point calculation unit 11 in the image data, the feature point extraction unit 31 sets one feature point as the feature point of interest, and sets feature points around the feature point of interest as A predetermined number of peripheral feature points are extracted in order from the closest distance from the feature point of interest. In the example of FIG. 8, when the predetermined number is 4 points and the feature point a is the feature point of interest, the feature points b, c, d, e are extracted as the peripheral feature points, and the feature point b is noted. In the case of feature points, four feature points a, c, e, and f are extracted as peripheral feature points.

  The feature point extraction unit 31 also extracts a combination of three points that can be selected from the four peripheral feature points extracted as described above. For example, as shown in FIGS. 9A to 9D, when the feature point a shown in FIG. 8 is the feature point of interest, three of the peripheral feature points b, c, d, and e are selected. Combinations, that is, combinations of peripheral feature points b, c, d, peripheral feature points b, c, e, peripheral feature points b, d, e, and peripheral feature points c, d, e are extracted.

  The invariant calculation unit 32 calculates, for each combination extracted by the feature point extraction unit 31, an invariant (one of feature amounts) Hij with respect to geometric deformation.

  Here, i is a number indicating the feature point of interest (i is an integer equal to or greater than 1), and j is a number indicating a combination of three peripheral feature points (j is an integer equal to or greater than 1). In the present embodiment, the ratio of two of the lengths of the line segments connecting the peripheral feature points is set as the invariant Hij.

  The length of the line segment can be calculated based on the coordinate value of each peripheral feature point. For example, in the example of FIG. 9A, if the length of the line segment connecting the feature point b and the feature point c is A11 and the length of the line segment connecting the feature point b and the feature point d is B11, The variable H11 is H11 = A11 / B11. In the example of FIG. 9B, if the length of the line segment connecting the feature point b and the feature point c is A12 and the length of the line segment connecting the feature point b and the feature point e is B12, The variable H12 is H12 = A12 / B12. In the example of FIG. 9C, if the length of the line segment connecting the feature point b and the feature point d is A13 and the length of the line segment connecting the feature point b and the feature point e is B13, The variable H13 is H13 = A13 / B13. In the example of FIG. 9D, if the length of the line segment connecting the feature point c and the feature point d is A14 and the length of the line segment connecting the feature point c and the feature point e is B14, The variable H14 is H14 = A14 / B14. In this way, invariants H11, H12, H13, and H14 are calculated in the examples of FIGS. 9A to 9D.

  In the above example, the line segment connecting the peripheral feature point closest to the target feature point and the second closest peripheral feature point is Aij, and the peripheral feature point closest to the target feature point and the third closest peripheral feature point are The line segment connecting the two is defined as Bij. However, the present invention is not limited to this, and the line segment used for calculating the invariant Hij may be selected by an arbitrary method.

The hash value calculation unit 33 is, for example,
Hi = (Hi1 × 10 ³ + Hi2 × 10 ² + Hi3 × 10 ¹ + Hi4 × 10 ⁰ ) / D
Is calculated as a hash value (one of feature quantities) Hi and stored in the obtained hash value memory 8. Note that D is a constant set in advance according to how much the range of values that the remainder can take is set.

  The method for calculating the invariant Hij is not particularly limited. For example, a multi-ratio of five points near the feature point of interest, a multi-ratio of five points extracted from n points in the vicinity (n is an integer of n ≧ 5), m points extracted from the n points in the vicinity (m is m <n and A value calculated based on the arrangement of m ≧ 5) and a cross ratio of five points extracted from m points may be used as the invariant Hij for the feature point of interest. The cross ratio is a value obtained from four points on a straight line or five points on a plane, and is known as an invariant with respect to projective deformation which is a kind of geometric transformation.

  Further, the formula for calculating the hash value Hi is not limited to the above formula, and other hash functions (for example, any one of the hash functions described in Patent Document 2) may be used.

  When the extraction of the peripheral feature points for one target feature point and the calculation of the hash value Hi are finished, each unit of the feature amount calculation unit 12 changes the target feature point to another feature point and extracts the peripheral feature points and the hash value. To calculate hash values for all feature points.

  In the example of FIG. 8, when the extraction of the peripheral feature points and the hash value when the feature point a is the target feature point is finished, the peripheral feature points and the hash value are extracted when the feature point b is the target feature point. I do. In the example of FIG. 8, when the feature point b is the target feature point, four feature points a, c, e, and f are extracted as the peripheral feature points.

  Then, as shown in FIGS. 10A to 10D, a combination of three points selected from these peripheral feature points a, c, e, and f (peripheral feature points a, e, f, and peripheral points) Feature points a, e, c, peripheral feature points a, f, c, and peripheral feature points e, f, c) are extracted, and a hash value Hi is calculated for each combination and stored in the memory 3. Then, this process is repeated for each feature point, and a hash value when each feature point is a feature point of interest is obtained and stored in the memory 3.

  In addition, when performing the registration process, the feature amount calculation unit 12 stores the hash value (feature amount) for each feature point of the input image data (image data of the registered image) calculated as described above. Send to.

  The registration processing unit 15 sequentially registers a hash value for each feature point calculated by the feature amount calculation unit 12 and an ID representing a registered image of the input image data in a hash table (not shown) provided in the memory 3. (See FIG. 11 (a)). If a hash value has already been registered, an ID is registered in association with the hash value. IDs are sequentially assigned numbers without duplication.

  If the number of registered images registered in the hash table exceeds a predetermined value (for example, 80% of the number of manuscripts that can be registered), old IDs may be searched and sequentially deleted. The erased ID may be used again as the ID of new input image data. Further, when the calculated hash values are the same value (H1 = H5 in the example of FIG. 11B), these may be combined and registered in the hash table.

  Further, when performing the collation processing, the feature amount calculation unit 12 sends a hash value for each feature point of the input image data (image data of the input image) calculated as described above to the voting processing unit 13.

  The voting processing unit 13 compares the hash value of each feature point calculated from the input image data with the hash value registered in the hash table, and votes for a registered image having the same hash value (see FIG. 12). FIG. 12 is a graph showing an example of the number of votes for three registered images ID1, ID2, and ID3. In other words, for each registered image, the voting processing unit 13 counts the number of times that the same hash value as the registered image has been calculated from the input image data, and stores the count value in the memory 3.

  Further, in the example of FIG. 11B, H1 = H5, and these are grouped into one of H1 and registered in the hash table. In such a table value, the input image calculated from the input image data is used. If H1 is included in the hash value of the registered image ID1, two votes are cast.

  The document similarity determination processing unit 14 extracts the registered image ID and the number of votes obtained from the voting process result of the voting processing unit 13 and obtains the extracted number of votes as a predetermined threshold value. The document similarity L1 is calculated by comparison, or the extracted number of votes is normalized by dividing by the maximum number of votes the document has, and the result is compared with a predetermined threshold value. The document similarity L1 is calculated.

  As an example of the threshold value in this case, for example, a method of setting to 0.8 or more can be mentioned. If there is a handwritten part, the number of votes may be greater than the maximum number of votes, so the similarity may be greater than one.

  The maximum number of votes is represented by the number of feature points × the number of hash values calculated from one feature point (attention feature point). In the example of FIGS. 8, 9, and 10 described above, as the simplest example, one hash value is calculated from one feature point. However, the feature points around the feature point of interest are selected. By changing the method, a plurality of hash values can be calculated from one feature point. For example, there are six combinations of extracting six points as feature points around the feature point of interest and extracting five points from these six points. Then, for each of these six ways, there is a method of calculating a hash value by extracting three points from five points to obtain an invariant.

  In the present embodiment, the document similarity L1 calculated by the document similarity determination processing unit 14 is output to the similarity total determination unit 5.

  Next, the color similarity determination unit 4 in the image matching device 101 will be described. As described above, the color similarity determination unit 4 includes a hue information extraction unit 41, a hue similarity determination processing unit 42, a saturation determination processing unit 43, and a hue information registration processing unit 44.

When the input image data, which is image data such as a registered image or an input image, is a color image, the hue information extraction unit 41 converts the RGB data of the input image data into a CIE 1976 L ^* a ^* b ^* signal (CIE: Commission International de l'Eclairage: International Lighting Commission, L ^* : Lightness, a ^* , b ^* : Chromaticity) For example, the hue information H is obtained from the following equation (2).

The process of extracting the hue information is not limited to the above method. For example, the CIE 1976 L ^* u ^* v ^* signal (CIE: Commission International de l'Eclairage: International Illumination Committee) is used instead of the L ^* a ^* b ^* signal. L ^* : Lightness index, u ^*, v ^* : Chromaticness index), hue information in each color space may be obtained, or other simple method for obtaining hue information May be used.

  The hue information extraction unit 41 creates, for example, a histogram (hereinafter referred to as a hue histogram) with 36 steps in increments of 10 (that is, a hue with a hue angle of every 36 degrees) using the calculated hue information.

  In the registration mode, the hue information registration processing unit 44 stores the hue histogram created by the hue information extraction unit 41 in the memory 3 in association with the ID representing the registered image. In the collation mode, the hue information registration processing unit 44 is passed through.

  In the collation mode, the hue similarity determination processing unit 42 compares the hue histogram created by the hue information extraction unit 41 with the hue histogram of the registered image to determine whether the input image and the registered image are similar. Is determined.

  At this time, the hue similarity determination processing unit 42 does not extract the hue histograms of all the registered images, but refers to the collation result of the document similarity determination processing unit 14 of the document collation unit 2 to obtain the shape feature quantity. The hue histogram is extracted for the registered images having similarity to.

  With reference to the collation result of the document similarity determination processing unit 14, for example, a process for specifying a registered image having similarity in the comparison of the shape feature amount is provided with a threshold and registration having a similarity equal to or higher than the threshold. All images may be extracted, or a threshold value and the number of extractions may be determined, and a predetermined number of registered images showing a high degree of similarity equal to or higher than the threshold value may be extracted.

  Then, the hue similarity determination processing unit 42 obtains the hue similarity using the histogram of the hue information as follows, for example.

  That is, when the hue similarity Lh is set, Lh is calculated from the hue histogram Hg = (b1, b2,... B36) of the input image and the hue histogram Hgr = (br1, br2,... Br36) of the registered image. The following equation (3) is used for calculation.

  The hue similarity Lh is considered to be higher as the value is smaller. The hue similarity Lh calculated by the hue similarity determination processing unit 42 is output to the similarity total determination unit 5.

  If the document similarity determination processing unit 14 determines that there is no similar registered image, the hue similarity determination processing unit 42 cancels the subsequent processing.

On the other hand, the hue information extraction unit 41 converts the signal into a CIE 1976 L ^* a ^* b ^* signal (CIE: Commission International de l'Eclairage: International Lighting Commission. L ^* : brightness, a ^* , b ^* : chromaticity). Image data such as registered images and input images is also sent to the saturation determination processing unit 43.

  The saturation determination processing unit 43 obtains the saturation information S from the CIE 1976 L * a * b * signal in the collation process using the following equation (4).

Note that the above is an example for obtaining the saturation information S. For example, an L ^* u ^* v ^* signal may be used, or the saturation is simply calculated from the RGB signal using the following equation (5). A value corresponding to may be obtained.

  For example, the saturation determination processing unit 43 obtains an average for all pixels of the input image from the calculated saturation information S, compares the obtained average value with a threshold (for example, 10% of the maximum saturation), and When the average value is smaller than the threshold value, the input image is determined to be “achromatic”. In the opposite case, it is determined as “chromatic”.

  Next, the similarity total determination unit 5 in the image collating apparatus 101 will be described. The similarity similarity determination unit 5 receives the document similarity L1 from the document similarity determination processing unit 14 of the document matching unit 2 and the hue similarity determination unit 42 of the color similarity determination unit 4. Lh is input from the saturation determination processing unit 43 of the color similarity determination unit 4 as “chromatic” or “achromatic”.

  For the document similarity L1 and the hue similarity Lh, the similarity similarity determination unit 5 determines that the similarity is high if the value is larger than a threshold value specified in advance, and the similarity is low if the value is low.

  For the document similarity L1, the threshold value for determining that the similarity is high is set to 90% of the maximum similarity, for example. On the other hand, in the hue similarity Lh, the threshold value for determining that the similarity is high is set to 90% of the maximum similarity, for example.

  For example, the similarity total determination unit 5 defines a relationship as shown in FIG. 13 and includes document similarity Ll, hue similarity Lh, and “chromatic” or “achromatic” information. In addition, the similarity is determined comprehensively and the similarity is determined.

  As can be seen from FIG. 13, when the document similarity L1 is low and the hue similarity Lh is high, and when the document similarity L1 is high and the hue similarity Lh is low, the document similarity L1 is high and the hue similarity Lh is low. In this case, it is determined that the similarity with the registered image is higher.

  By giving priority to the similarity determination result based on the shape feature amount of the image over the similarity determination result based on the color feature amount of the image, the similarity between the registered image and the input image is more accurate than prioritizing the reverse. Can be determined.

  Further, when there are two images having the same document similarity L1 and the same low hue similarity Lh, the similarity comprehensive determination unit 5 is achromatic based on the result of the saturation determination of the input image. The case is determined to be higher than the case of being chromatic.

  Here, the reason why the case where the input image is achromatic in this way is determined to be higher than chromatic will be described.

  When judging the similarity between the registered image and the input document comprehensively by combining the judgment based on the shape feature quantity and the judgment based on the color feature quantity, the similarity based on the shape feature quantity is about the same and the similarity based on the hue is about the same. It is determined that the two images showing low values are not similar to the registered image.

  For example, for example, an image composed of green characters and ruled lines and a black character determined to be similar in the determination by the shape feature amount to an image composed of red characters and ruled lines. The two images of the ruled line image are determined not to be similar to the image of the red color character or ruled line in the similarity determination by hue.

  However, in practice, an original image of a color image is often copied as a monochrome image or transmitted by facsimile and used as an achromatic non-original image. In addition, as in the application form described above, there are many document images in which characters of different colors and ruled lines are used for document images having a similar layout (high similarity due to shape feature amounts). Furthermore, it is almost unthinkable to convert a color document from its original color to another chromatic color.

  Therefore, in the previous example, it is better to determine that the similarity of a black text / ruled line document is higher than that of a green text / ruled line document in consideration of the usage status of the document image. Become.

  Therefore, in the present embodiment, as described above, the similarity total determination unit 5 determines the saturation of the input image when there are two images having the same document similarity Ll and the same low hue similarity Lh. As a result of the degree determination, the case where the color is achromatic is determined to be higher than the case where the color is chromatic.

  As a result, an achromatic non-original image formed by copying the original color image as described above into a black-and-white copy or facsimile transmission, or characters and ruled lines of similar colors but different in layout are used. With respect to an image such as the application form described above, it is possible to make a determination in accordance with the user's sense, and a highly accurate determination is possible.

  In addition to providing a table that defines the relationship as shown in FIG. 13, an expression for calculating a similarity that is integrated in advance from the two similarities (document similarity L1 and hue similarity Lh) is provided. It may be determined and calculated. In that case, when the value of the hue similarity Lh is lower than a predetermined threshold value, the result of saturation determination may be referred to correct so that the similarity is higher in the achromatic case.

  Next, the configuration of the digital color copying machine 102 provided with the image collating apparatus 101 will be described. FIG. 2 is a block diagram showing the configuration of the digital color copying machine 102.

  As shown in FIG. 2, the digital color copying machine 102 includes a color image input device 111, a color image processing device 112, a color image output device 113, and an operation panel 114.

  The color image input device 111 includes a scanner unit including a device that converts optical information such as a CCD (Charge Coupled Device) into an electrical signal, and outputs a reflected light image from the original as an RGB analog signal.

  An analog signal read by the color image input device 111 is converted into an A / D conversion unit 121, a shading correction unit 122, an image matching unit 123, an input tone correction unit 124, and a region separation processing unit in the color image processing device 112. 125, a color correction unit 126, a black generation and under color removal unit 127, a spatial filter processing unit 128, an output tone correction unit 129, and a tone reproduction processing unit 130 are sent in this order, and a color image is obtained as a CMYK digital color signal. It is output to the output device 113.

  The A / D conversion unit 121 converts RGB analog signals into digital signals. The shading correction unit 122 applies a color image input device to the digital RGB signals sent from the A / D conversion unit 121. Processing for removing various distortions generated in the illumination system 111, imaging system, and imaging system 111 is performed. Further, at the same time as adjusting the color balance, a process for converting the signal such as a density signal into an easy-to-handle signal of the image processing system employed in the color image processing apparatus 112 is performed.

  The image matching unit 123 determines the degree of similarity between the input image data input and the registered image registered in advance, and outputs the result, and corresponds to the image matching device 101 in FIG. Further, the image matching unit 123 outputs the input RGB data of the image data as it is to the input tone correction unit 124 at the subsequent stage. In addition, the image matching unit 123 outputs the obtained similarity determination result (determination signal), and in response to this, the operation panel 114 displays a list of images determined to be similar, or the subsequent processing unit performs processing. Or change. For example, copying, electronic distribution, filing prohibition, electronic distribution to a predetermined address, filing to a folder, and the like are performed.

  The input tone correction unit 124 performs image quality adjustment processing such as removal of background density and contrast on the RGB signal from which various distortions are removed by the shading correction unit 122.

  The region separation processing unit 125 separates each pixel in the input image into one of a character region, a halftone dot region, and a photographic region from the RGB signal. The region separation processing unit 125 sends a region identification signal indicating which region the pixel belongs to to the black generation and under color removal unit 127, the spatial filter processing unit 128, and the gradation reproduction processing unit 130 based on the separation result. In addition to the output, the input signal output from the input tone correction unit 124 is directly output to the subsequent color correction unit 126.

  The color correction unit 126 performs a process of removing color turbidity based on the spectral characteristics of the CMY color material including unnecessary absorption components in order to make color reproduction faithful.

  The black generation and under color removal unit 127 generates black (K) signals from the CMY three-color signals after color correction, and subtracts the K signals obtained by black generation from the original CMY signals to generate new CMY signals. The process to generate is performed. As a result, the CMY three-color signal is converted into a CMYK four-color signal.

  The spatial filter processing unit 128 performs spatial filter processing using a digital filter on the image data of the CMYK signal input from the black generation and under color removal unit 127 to correct the spatial frequency characteristics. As a result, blurring of the output image and deterioration of graininess can be reduced.

  Similar to the spatial filter processing unit, the gradation reproduction processing unit 129 performs predetermined processing described later on the image data of the CMYK signal based on the region identification signal.

  For example, in the region separated into characters by the region separation processing unit 125, a filter having a high enhancement amount of a high frequency component is used as a spatial filter in the spatial filter processing unit 128 in order to improve the reproducibility of characters. At the same time, the gradation reproduction processing unit 130 performs binarization or multi-value processing using a high-resolution screen suitable for reproducing high-frequency components.

  In addition, with respect to the region separated into halftone dots by the region separation processing unit 125, the spatial filter processing unit 128 performs low-pass filter processing for removing the input halftone component. The output tone correction unit 129 performs output tone correction processing for converting a signal such as a density signal into a halftone dot area ratio that is a characteristic value of the color image output device 113, and then the tone reproduction processing unit 130. A gradation reproduction process is performed in which the image is finally separated into pixels and each gradation is reproduced. With respect to the region separated into photographs by the region separation processing unit 125, binarization or multi-value processing is performed on the screen with an emphasis on gradation reproducibility.

  The image data subjected to the above-described processes is temporarily stored in a storage device (not shown), read out at a predetermined timing, and input to the color image output device 113.

  The color image output device 113 outputs image data on a recording medium such as paper. For example, a color imageless output device using an electrophotographic method or an ink jet method can be used, but the color image output device 113 is particularly limited. It is not a thing. The above processing is controlled by a CPU (Central Processing Unit) not shown.

  In the above configuration, the operation of the image collating apparatus 101 of the present embodiment will be described below based on the flowcharts of FIGS.

  FIG. 14 is a flowchart in the registration mode. The registration mode is selected by operating the operation panel 114 in the digital color copying machine 102. In an image processing system including the image collating apparatus 101 and a terminal device (computer) connected thereto, the image is selected by an input operation from the terminal device.

  When image data of a registered image to be registered is input, the control unit 1 extracts and registers a shape feature amount, calculates hue information that is a color feature amount, creates a hue histogram, and registers it The processing to be performed is performed in parallel. In addition, it is good also as a structure which processes not only in parallel processing but in order.

  In the process of extracting and registering shape feature amounts, the feature point calculation unit 11 calculates each feature point of the registered image based on the input image data (S1), and the feature amount calculation unit 12 The feature amount of each feature point calculated by the point calculation unit 11 is calculated (S2), and the registration processing unit 15 represents the feature amount (hash value) of the feature point of the registered image and the registered image. The index (ID) is stored in the table of the memory 3 (S3).

  On the other hand, in the process of creating and registering a hue histogram by extracting the hue information that is the color feature amount, the hue information extraction unit 41 extracts the hue information of the registered image based on the input image data, and the hue A histogram is created (S4), and the hue information registration processing unit 44 stores the hue histogram of the registered image and an index (ID) representing the registered image in the table of the memory 3 (S5). When the above processing is performed, the control unit 1 ends the registration mode.

  FIG. 15 is a flowchart in the collation mode. The processing mode is entered when a document image is read by the image collating apparatus 101, such as copying, facsimile transmission, or filing, and some processing is to be performed.

  When the image data of the input image to be processed is input, the control unit 1 extracts a shape feature amount and performs collation, calculates hue information calculation that is a color feature amount, creates a hue histogram, and collates The processing to be performed is performed in parallel. In addition, it is good also as a structure which processes not only in parallel processing but in order.

  In the process of extracting and collating the shape feature amount, the feature point calculation unit 11 calculates each feature point of the input image based on the input image data (S11), and the feature amount calculation unit 12 calculates the feature point. The feature amount of each feature point calculated by the unit 11 is calculated (S12), and the voting processing unit 13 performs a voting process using the calculated feature amount of the input image (S13).

  Next, the document similarity determination processing unit 14 determines whether the input image is similar to any registered image based on the result of the voting process (S14). If the registered image is not similar to any registered image, the document similarity determination processing unit 14 outputs a determination signal “0” (S15). On the other hand, when similar to any registered image, the document similarity LI is output (S16).

  On the other hand, in the process of extracting the hue information, which is the color feature amount, and creating and comparing the hue histogram, the hue information extraction unit 41 extracts the hue information of the input image based on the input image data, and the hue histogram is obtained. Then, the hue similarity determination processing unit 42 compares the created hue histogram of the input image with the hue histogram of the registered image acquired from the table of the memory 3 to calculate the hue similarity (S22). ). Here, the hue similarity determination processing unit 42 refers to the determination result of the document similarity determination processing unit 14 and extracts a hue histogram of a registered document that is determined to be similar in the determination based on the shape feature amount.

  Next, the saturation determination processing unit 43 determines whether the input image is chromatic or achromatic based on the input image data (S23), and the hue similarity determination processing unit 42 and the saturation determination processing unit. 43 outputs the determination result of hue similarity and saturation (S24).

  The similarity total determination unit 5 comprehensively determines the similarity between the input image and the registered image based on the input document similarity Ll and hue similarity Lh, and chromatic and achromatic information ( S25), a determination signal is output (S26). In the present embodiment, the similarity comprehensive determination unit 5 determines the similarity as a comprehensive determination based on the input document similarity Ll and hue similarity Lh, and chromatic and achromatic information. Since it is shown in 8 stages, this similarity is output as a determination signal.

  As described above, the image matching apparatus 101 according to the present embodiment extracts a shape feature amount indicating a feature related to the shape of an image from input image data, and compares it with the shape feature amount of a registered image registered. Determining the similarity between the input image and the registered image, extracting a color feature amount indicating the color feature of the image from the input image data, and comparing it with the color feature amount of the registered image; The similarity between the input image and the registered image is determined, and the saturation of the input image is extracted and the saturation information is added to determine the similarity between the input image and the registered image. It has become.

  As a result, even if the input image and the registered image are color images such as application forms that cannot be identified by the similarity determination based on the shape feature amount, and the components having different colors of characters and ruled lines are similar, Similarity can be determined with high accuracy.

  Similarly, the input image and the registered image are color images with similar colors that could not be identified by the similarity determination based on the color feature amount, or the registered image is a color image and the input image is black and white. Even for black and white images formed by copying or facsimile transmission, similarity can be determined with high accuracy.

  Then, by determining using the saturation information in addition to the component of the image and the color information, the red ruled line document having the same degree of similarity with respect to the shape (character and ruled line shape) described above can be obtained. For three sheets, a green-formed document and a black ruled document, 1) a red ruled document and a green ruled document, and 2) a red ruled document and a black ruled document than in 1) It is possible to carry out a determination in accordance with the actual usage such that 2) has a higher degree of similarity.

  Thereby, when the original image is a color image, it is possible to accurately determine the similarity to a non-original image formed by copying the original image.

  Such similarity determination is used for image search so that images are sorted and displayed according to the order of similarity obtained, whereby an original color image can be copied in black and white or facsimile. The transmitted images can be displayed in a higher order than completely different color originals having similar shape characteristics.

  FIG. 16 is a block diagram showing a configuration of a digital color multifunction peripheral (image data output processing apparatus) 104 provided with the image collating apparatus 101 of this embodiment.

  The digital color multifunction peripheral 104 has a configuration in which a communication device 115 such as a modem or a network card is added to the digital color copying machine 102 shown in FIG.

  In the digital color multifunction peripheral 104, when transmitting a facsimile, the communication device 115 performs a transmission procedure with the other party, and when a state where transmission is possible is secured, image data compressed in a predetermined format (scanner) Is read from the memory 3 and necessary processing such as changing the compression format is performed, and the image data is sequentially transmitted to the other party via the communication line.

  When receiving a facsimile, the image data transmitted from the other party is received and input to the color image processing apparatus 116 while performing a communication procedure. In the color image processing apparatus 116, the received image data is expanded by a compression / expansion processing unit (not shown). The decompressed image data is subjected to rotation processing and resolution conversion processing as necessary, and output tone correction (output tone correction unit 131) and tone reproduction processing (tone reproduction processing unit 132) are performed. Are output from the color image output device 113.

  The digital color multifunction peripheral 104 performs data communication with a computer and other digital multifunction peripherals connected to the network via a network card and a LAN cable.

  In the above example, the digital color multifunction peripheral 104 has been described. However, the multifunction peripheral may be a monochrome multifunction peripheral. A single facsimile communication apparatus may be used.

  Further, the image collating apparatus 101 of the present embodiment can also be applied to an image reading apparatus. FIG. 17 is a block diagram showing the configuration of a color image reading apparatus (image data output processing apparatus) 105 to which the image collating apparatus 101 of this embodiment is applied. The color image reading device 105 is a flatbed scanner, for example, and may be a digital camera.

  The color image reading device 105 includes a color image input device 111 and a color image processing device 117, and the color image processing device 117 includes an A / D conversion unit 121, a shading correction unit 122, and an image matching unit 123. The image matching unit 123 'corresponds to the image matching apparatus 101 shown in FIG.

  The color image input device 111 (image reading means) is composed of, for example, a scanner unit equipped with a CCD (Charge Coupled Device), and the reflected light image from the original is converted into RGB (R: red, G: green, B: blue). The analog signal is read by the CCD and input to the color image processing device 117.

  The analog signal read by the color image input device 111 is sent through the color image processing device 117 in the order of an A / D (analog / digital) conversion unit 121, a shading correction unit 122, and an image matching unit 123 '.

  The A / D converter 121 converts RGB analog signals into digital signals, and the shading correction unit 122 applies a color image input device to the digital RGB signals sent from the A / D converter 121. Processing for removing various distortions generated in the illumination system 111, imaging system, and imaging system 111 is performed. The shading correction unit 122 adjusts the color balance and converts the RGB reflectance signal into a density signal.

  The function of the image matching unit 123 ′ is as described above in the image matching device 101. The image collating unit 123 ′ determines the similarity between the input image data and the registered image, and the result (determination signal: prohibition of copying, electronic distribution, filing, electronic distribution to a predetermined address, filing to a folder, etc. Output). The determination signal is transmitted together with the read image data to a printer or multifunction device via a network and output. Or it inputs into a printer via a computer or directly into a printer. In this case, it is necessary to be able to determine a signal representing the processing contents on the printer, the multifunction peripheral, or the computer side. Rather than outputting the determination signal, the calculated feature value, color feature value (hue information), and saturation of the input image are output, and the server, the computer, or the printer checks the registered image with the registered image. May be performed. A digital camera may be used as the image reading device.

  The present invention performs the above-described similarity determination (image collation) and output control on a computer-readable recording medium in which program codes (executable program, intermediate code program, source program) of a program to be executed by a computer are recorded. An image processing method may be recorded. As a result, the recording medium on which the program code for performing the similarity determination unit, the output control, and the image processing method for registering the original image is recorded can be provided in a portable manner.

  In this embodiment, since the processing is performed by a microcomputer, the recording medium may be a memory (not shown), for example, a ROM itself as a program medium, or an external storage device (see FIG. (Not shown) may be a program medium provided with a program reading device, which can be read by inserting a recording medium therein.

  In any case, the stored program may be configured to be accessed and executed by the microprocessor, or in any case, the program code is read and the read program code is stored in the microcomputer. The program code may be downloaded to a program storage area (not shown) and the program code is executed. It is assumed that this download program is stored in the main device in advance.

  Here, the program medium is a recording medium configured to be separable from the main body, such as a tape system such as a magnetic tape or a cassette tape, a magnetic disk such as a floppy (registered trademark) disk or a hard disk, or a CD-ROM / MO /. MD / DVD optical discs, IC cards (including memory cards) / optical cards, etc. Mask ROM, EPROM (Erasable Programmable Read Only Memory), EEPROM (Electrically Erasable Programmable Read Only Memory), flash A medium that carries a fixed program including a semiconductor memory such as a ROM may be used.

  In the present embodiment, since the system configuration is such that a communication network including the Internet can be connected, it may be a medium that fluidly carries a program so as to download a program code from the communication network. When the program code is downloaded from the communication network in this way, the download program may be stored in the main device in advance, or may be installed from another recording medium. The present invention can also be realized in the form of a computer data signal embedded in a carrier wave in which the program code is embodied by electronic transmission.

  The recording medium is read by a program reading device provided in a digital color image forming apparatus or a computer system, whereby the above-described image processing method is executed.

  The computer system also displays an image input device such as a flatbed scanner, a film scanner, and a digital camera, a computer that performs various processes such as the above image processing method by loading a predetermined program, and displays the processing results of the computer. It comprises an image display device such as a CRT display / liquid crystal display and a printer that outputs the processing results of the computer to paper or the like. Furthermore, a network card, a modem, and the like are provided as communication means for connecting to a server or the like via a network.

  The present invention is not limited to the above-described embodiments, and various modifications are possible within the scope shown in the claims, and embodiments obtained by appropriately combining technical means disclosed in different embodiments. Is also included in the technical scope of the present invention.

It is a block diagram which shows the structure of the image collation apparatus of embodiment of this invention. FIG. 2 is a block diagram illustrating a configuration of a digital color copying machine that is an image data output processing apparatus including the image collating apparatus illustrated in FIG. 1. It is a block diagram which shows the structure of the feature point calculation part in the image collation apparatus shown in FIG. It is explanatory drawing which shows the filter coefficient of the mixing filter with which the MTF process part in the feature point calculation part shown in FIG. 3 is provided. It is explanatory drawing which shows an example of the connection area | region extracted from the image data binarized by the process of the feature point calculation part shown in FIG. 3, and the gravity center of this connection area | region. It is explanatory drawing which shows an example of each gravity center (feature point) of the some connection area | region extracted from the character string contained in the binarized image data by the process of the feature point calculation part shown in FIG. It is a block diagram which shows the structure of the feature-value calculation part in the image collation apparatus shown in FIG. It is explanatory drawing of the extraction operation | movement of the surrounding feature point with respect to the attention feature point in the feature point extraction part in the feature-value calculation part shown in FIG. FIG. 9A shows an example of a combination of three points that can be selected from the four peripheral feature points extracted by the feature point extraction unit shown in FIG. FIG. 9B is an explanatory diagram illustrating an example of a combination of peripheral feature points b, c, and d. FIG. 9B is an explanatory diagram illustrating an example of a combination of peripheral feature points b, c, and e with respect to the target feature point a. FIG. 9C is an explanatory diagram showing an example of a combination of peripheral feature points b, d, and e with respect to the target feature point a, and FIG. 9D is a combination of peripheral feature points c, d, and e with respect to the target feature point a. It is explanatory drawing which shows the example of. FIG. 10A shows three peripheral feature points that can be selected when the target feature point is transferred to one of the four peripheral feature points extracted by the feature point extraction unit shown in FIG. FIG. 10B is a diagram illustrating an example of the combination, and is an explanatory diagram illustrating an example of the combination of the peripheral feature points a, e, and f with respect to the target feature point b. FIG. FIG. 10C is an explanatory diagram illustrating an example of a combination of peripheral feature points a, f, and c with respect to the feature point of interest b, and FIG. It is explanatory drawing which shows the example of the combination of the surrounding feature points e, f, and c with respect to the attention feature point b. FIGS. 11A and 11B are explanatory diagrams illustrating an example of a hash value and a registered image index for each feature point stored in the memory in the image collating apparatus illustrated in FIG. 1. It is a graph which shows an example of the voting result by the voting process part in the image collation apparatus shown in FIG. Relationship between shape similarity Ll, hue similarity Lh, and “chromatic” or “achromatic” information used by the similarity similarity determination unit 5 in the image matching apparatus shown in FIG. 1 and the overall similarity It is explanatory drawing which shows the content of the table which prescribed | regulated. It is a flowchart which shows operation | movement of the registration mode of the image collation apparatus shown in FIG. It is a flowchart which shows operation | movement of the collation mode of the image collation apparatus shown in FIG. FIG. 2 is a block diagram illustrating a configuration of a digital color multifunction peripheral that is an image data output processing apparatus including the image collating apparatus of FIG. 1. It is a block diagram which shows the structure of the color image reading apparatus which is an image data output processing apparatus provided with the image collation apparatus of FIG.

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 Control part 2 Document collation part (shape similarity determination part)
3 Memory (memory means)
4 color similarity determination unit 11 feature point calculation unit 12 feature amount calculation unit 13 voting processing unit 14 document similarity determination processing unit 15 registration processing unit 101 image collation device 102 digital color copying machine (image data output processing device)
103 Image Processing Device 104 Digital Color Multifunction Device (Image Data Output Processing Device)
105 Color image reading device (image data output processing device)
112 Color Image Processing Device 116 Color Image Processing Device 117 Color Image Processing Device 123 Image Collation Unit (Image Collation Device)

Claims (11)

A shape feature amount indicating a feature related to the shape of the image is extracted from the input image data, compared with the shape feature amount of the registered image, and the similarity between the input image and the registered image is determined. A shape similarity determination unit for determining;
Color similarity that extracts the feature of the image color from the input image data, compares it with the color feature of the registered image, and determines the similarity between the input image and the registered image A determination unit;
A saturation determination unit that extracts saturation from the input image data and determines whether the input image is chromatic or achromatic;
Comprehensive similarity determination for determining the similarity between the input image and the registered image based on the determination result of the shape similarity determination unit, the determination result of the color similarity determination unit, and the determination result of the saturation determination unit And an image collating apparatus. The color similarity determination unit
Using hue as the color feature amount,
A hue extraction unit that extracts the hue from the input image data;
A hue similarity determining unit that compares the hue of the input image extracted by the hue extracting unit with the hue of the registered image to determine the similarity between the input image and the registered image. Item 2. The image matching device according to Item 1.   When the similarity based on the shape feature amount is similar and the similarity based on the color feature amount is low, the similarity comprehensive determination unit determines that the input image is achromatic and chromatic, and the achromatic one is the registered image. The image collating apparatus according to claim 1, wherein the similarity is determined to be high.   The similarity total determination unit determines whether the shape feature is low when the similarity by the shape feature amount is low and the similarity by the color feature amount is high, and when the similarity by the shape feature amount is high and the similarity by the color feature amount is low. The image collating apparatus according to claim 1, 2 or 3, wherein the similarity with the amount is high and the similarity with the color feature amount is low, the similarity with the registered image is determined to be high.   In extracting the color feature amount, the color similarity determination unit refers to the determination result by the shape similarity determination unit, and only for the registered image determined to have high similarity by the shape feature amount. 5. The image collating apparatus according to claim 1, wherein: The shape similarity determination unit
A feature point calculation unit that calculates a feature point by extracting a connected portion of the input image from input image data;
A shape feature amount calculation unit that calculates a shape feature amount of the input image based on a relative position between the feature points calculated by the feature point calculation unit;
A shape similarity calculation unit that compares the shape feature value of the input image calculated by the shape feature value calculation unit with the shape feature value of the registered image and calculates the similarity of the shape of the input image and the registered image The image collating apparatus according to claim 1, 2, 3, 4 or 5. Shape similarity in which similarity between the first image and the second image is determined by comparing shape feature amounts indicating features related to the shape of the image extracted from the image data of the first image and the second image. A determination unit;
Color similarity for determining the similarity between the first image and the second image by comparing color feature amounts indicating characteristics relating to the colors of the images extracted from the image data of the first image and the second image A determination unit;
A saturation determination unit that extracts saturation from each image data of the first image and the second image, and determines whether the first image and the second image are chromatic or achromatic;
The similarity between the first image and the second image is determined based on the determination result of the shape similarity determination unit, the determination result of the color similarity determination unit, and the determination result of the saturation determination unit. An image collation apparatus comprising: a comprehensive similarity determination unit. An image data output processing device that performs output processing on image data,
An image data output processing device comprising the image collating device according to claim 1. A shape feature amount indicating a feature related to the shape of the image is extracted from the input image data, compared with the shape feature amount of the registered image, and the similarity between the input image and the registered image is determined. A shape similarity determination step for determining;
Color similarity that extracts the feature of the image color from the input image data, compares it with the color feature of the registered image, and determines the similarity between the input image and the registered image A determination process;
A saturation determination step for extracting saturation from the input image data and determining whether the input image is chromatic or achromatic;
Comprehensive similarity determination that determines the similarity between the input image and the registered image based on the determination result of the shape similarity determination step, the determination result of the color similarity determination step, and the determination result of the saturation determination step And an image collating method.   The program for functioning a computer as said each part of the image collation apparatus of any one of Claims 1-7.   The computer-readable recording medium which recorded the program of Claim 10.

Images