JP2006023834A - Image processor and its control method, and program - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide an image processor and its control method, and a program for executing the similar image retrieval of images including a document image with high precision. <P>SOLUTION: A comparison origin image is divided into a plurality of blocks. Whether or not any character image is present in each of the divided block images is decided on the basis of the image characteristic information of the divided block images. When it is decided that any character image is present in either of the block image and the block image of the comparison destination image corresponding to the block image, predetermined similarity is acquired as the similarity of them, and when it is decided that any character image is not present in either the block image and the block image of the comparison destination image corresponding to the block image, the similarity based on the image characteristic information of them is acquired. Then, an image being a retrieval result is outputted on the basis of cumulative similarity to be acquired by cumulatively adding the similarity to each block image. <P>COPYRIGHT: (C)2006,JPO&NCIPI

Description

  The present invention relates to an image processing apparatus that performs a similar image search based on the similarity between a comparison source image serving as a search condition and a comparison destination image, a control method thereof, and a program.

Conventionally, in a similar image search process, an image is divided into a plurality of vertical and horizontal blocks in a registration process, block representative colors such as an average color and a most frequent color for the block are calculated, and stored in association with the image. Keep it. In the search process, a distance calculation indicating the similarity is performed by accumulating the differences in the representative colors of the corresponding divided blocks, and based on the calculation result, the similar images are sorted and output. (For example, see Patent Document 1).
JP-A-7-239856

  The similar image search method as described above was useful for an image in which color information exists in the entire image such as a natural image, but an image is pasted on a white background like a document image, and characters are placed in the margin. If there is a character on the white background, even if there is a character on the white background, the average color or the most frequent color will not be affected, and it is not possible to distinguish between the divided block with the character and the white background, The search result was uncomfortable for the user. In particular, the reason why the average color or the most frequent color has no effect is that the ratio of the actual character pixels in the character image is small.

  Therefore, in order to compensate for such a problem, there is a method for obtaining the characteristics of the separated objects by using a technique for separating the document image into a character area and an image area in advance. According to this method, although the search accuracy is improved, there is a possibility that a desired image cannot be searched if the character area or the image area is not extracted.

  The present invention has been made to solve the above-described problem, and an object thereof is to provide an image processing apparatus, a control method thereof, and a program capable of executing a similar image search of an image including a document image with higher accuracy. And

In order to achieve the above object, an image processing apparatus according to the present invention comprises the following arrangement. That is,
An image processing apparatus that performs a similar image search based on a similarity between a comparison source image serving as a search condition and a comparison destination image,
Dividing means for dividing the comparison source image into a plurality of blocks;
Determination means for determining whether a character image exists in the block image based on image feature information of the block image divided by the dividing means;
As a result of the determination by the determination means, when a character image exists in either the block image or the block image corresponding to the comparison target image, a predetermined similarity is acquired as the similarity between the two, while the block image And when there is no character image in either of the block images of the comparison target image corresponding thereto, acquisition means for acquiring the similarity based on the image feature information of both,
Output means for outputting an image as a search result based on a cumulative similarity obtained by cumulatively adding similarities to the respective block images divided by the dividing means acquired by the acquiring means.

Preferably, the comparison unit includes a management table for managing a similarity according to image feature information between the comparison source image and the comparison destination image,
The management table includes the comparison source as a predetermined similarity based on image feature information between the comparison source image and the comparison destination image when a character image exists in either the comparison source image or the comparison destination image. A fixed value that is sufficiently different from the similarity based on the image feature information between the comparison source image and the comparison target image when no character image exists in the image is managed.

  Preferably, the determination unit determines whether or not an image feature amount indicated by the image feature information of the block image is equal to or greater than a predetermined value. If the image feature amount is equal to or greater than a predetermined value, the processing is performed. It is determined that a character image exists in the target image.

  Preferably, the image feature information is luminance feature information related to the luminance of the block image.

  Preferably, the luminance feature information is an average luminance or a mode luminance of the block image.

  Preferably, the luminance feature information is an average luminance or a mode luminance in a color vicinity range of a pixel having the highest luminance in the block image.

In order to achieve the above object, an image processing apparatus according to the present invention comprises the following arrangement. That is,
An image processing apparatus that performs image registration processing of an image to be searched for similar image search,
An input means for inputting an image;
Dividing means for dividing the image into a plurality of blocks;
Determination means for determining whether or not a character image exists in each block image based on image feature information of each block image divided by the dividing means;
If a character image exists in the block image as a result of the determination by the determining means, information indicating the character image is stored as image feature information. On the other hand, if a character image does not exist in the block image, the image feature of the block image is stored. Storage means for storing information.

In order to achieve the above object, a method for controlling an image processing apparatus according to the present invention comprises the following arrangement. That is,
A control method of an image processing apparatus that performs a similar image search based on a similarity between a comparison source image serving as a search condition and a comparison target image,
A dividing step of dividing the comparison source image into a plurality of blocks;
A determination step of determining whether or not a character image exists in the block image based on image feature information of the block image divided in the division step;
As a result of the determination in the determination step, when a character image exists in either the block image or the block image of the comparison target image corresponding thereto, a predetermined similarity is obtained as the similarity between the two, while the block And when there is no character image in either of the block images of the comparison target image corresponding thereto, an acquisition step of acquiring the similarity based on the image feature information of both,
An output step of outputting an image as a search result based on the cumulative similarity obtained by cumulatively adding the similarity to each block image divided in the division step acquired in the acquisition step.

In order to achieve the above object, a method for controlling an image processing apparatus according to the present invention comprises the following arrangement. That is,
A control method of an image processing apparatus that performs image registration processing of an image to be searched for similar image search,
An input process for inputting an image;
A dividing step of dividing the image into a plurality of blocks;
A determination step of determining whether or not a character image exists in each block image based on image feature information of each block image divided in the division step;
As a result of the determination in the determination step, if a character image exists in the block image, information indicating it is stored as image feature information. On the other hand, if a character image does not exist in the block image, the image feature of the block image A storage step of storing information in the storage unit.

In order to achieve the above object, a program according to the present invention comprises the following arrangement. That is,
A program that realizes control of an image processing apparatus that performs a similar image search based on a similarity between a comparison source image that is a search condition and a comparison target image,
A program code of a dividing step of dividing the comparison source image into a plurality of blocks;
Based on the image feature information of the block image divided in the division step, a program code of a determination step for determining whether or not a character image exists in the block image;
As a result of the determination in the determination step, when a character image exists in either the block image or the block image of the comparison target image corresponding thereto, a predetermined similarity is obtained as the similarity between the two, while the block And when there is no character image in either of the block images of the comparison target image corresponding thereto, the program code of the obtaining step for obtaining the similarity based on the image feature information of both,
Program code of an output step for outputting an image as a search result based on the cumulative similarity obtained by cumulatively adding the similarity to each block image divided in the division step acquired in the acquisition step.

In order to achieve the above object, a program according to the present invention comprises the following arrangement. That is,
A program that realizes control of an image processing apparatus that performs image registration processing of an image to be searched for similar image search,
An input process for inputting an image;
A dividing step of dividing the image into a plurality of blocks;
A determination step of determining whether or not a character image exists in each block image based on image feature information of each block image divided in the division step;
As a result of the determination in the determination step, if a character image exists in the block image, information indicating it is stored as image feature information. On the other hand, if a character image does not exist in the block image, the image feature of the block image A storage step of storing information in the storage unit.

  As described above, according to the present invention, it is possible to provide an image processing apparatus, a control method thereof, and a program capable of executing a similar image search of an image including a document image with higher accuracy.

  Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings.

  FIG. 1 is a block diagram of an image processing apparatus according to an embodiment of the present invention.

  In the present invention, an image registration process for registering an image and an image search process for searching for a desired image from a registered image group will be described in order.

  The image registration process is a process of inputting an image to be registered, extracting an image feature amount of the input image, and storing the input image and the image feature amount in association with each other.

  The image search process is a so-called similar image search process. Extraction of image feature amounts of a comparison source image serving as a search condition (or an image feature corresponding to the image when a registered image is used as a comparison source image) The image feature value is compared with the image feature value of the registered comparison target image, and an image similar to the comparison source image is searched based on the comparison result.

  In this embodiment, the image feature amount extraction at the time of the image registration process is performed by dividing the processing target image into a plurality of regions and extracting the image feature amount for each region. It is the same regardless of the aspect ratio and size of the processing target image. Similarly, the extraction of the image feature amount during the image search process is performed under the same conditions as the extraction of the image feature amount during the image registration process. The image feature amount here includes both color feature information relating to the color of the image and luminance feature information relating to the luminance of the image.

  In FIG. 1, reference numeral 101 denotes a user interface (UI) unit, which is a user interface for executing various processes including an image registration process and an image search process according to the present invention. This user interface is realized by, for example, a graphic user interface and an input device. The user can appropriately execute the image registration process and the image search process by using the UI unit 101.

  An image input processing unit 102 inputs an image to be registered by the image registration process. An image accumulation processing unit 103 accumulates the input image in the image accumulation unit 103a. Reference numeral 104 denotes an image management information processing unit that generates management information for managing the input image and stores it in the image management DB 104a. A luminance feature information extraction processing unit 105 extracts luminance feature information related to the luminance of the input image, and registers the extracted luminance feature information in the luminance feature information index table 105a. A color feature information extraction processing unit 106 extracts color feature information related to the color of the input image and registers the extracted color feature information in the color feature information index table 106a.

  Reference numeral 107 denotes a comparison source image input unit processing unit, which inputs an image (a comparison source image in a similar image search) that is a search condition by the image search process. A determination processing unit 108 determines whether or not the input color information of the comparison source image is sufficient.

  A color feature information comparison processing unit 109 performs similarity comparison between the comparison source image and the comparison destination image based on the color feature information. A luminance feature information comparison processing unit 110 performs similarity comparison between the comparison source image and the comparison destination image based on the luminance feature information.

  A search result display processing unit 111 displays an image as a search result based on the processing result of the color feature information comparison processing unit 109 or the luminance feature information comparison processing unit 110. In addition, the search result display processing unit 111 may be configured to print an image as a search result by a printer. In particular, when there is only one image as a search result and it is desired to print the image, the processing can be speeded up.

  Note that details of processes executed by various components constituting the image processing apparatus will be described in the following order.

  The image processing apparatus has standard components (for example, a CPU, a memory (RAM, ROM), a hard disk, an external storage device, a network interface, a display, a keyboard, a mouse, etc.) mounted on a general-purpose computer. Yes.

  Further, all or some of the various components shown in FIG. 1 may be realized by dedicated hardware, or may be realized by software executed under the control of the CPU.

  First, a case where image registration processing is executed via the UI unit 101 will be described.

[Image registration process]
In the image registration process, first, the image to be registered is input by the image input processing unit 102. The registration target image is input using, for example, an input device such as a scanner or by reading an image stored in an external storage device. The input image is temporarily stored in the memory, and from the data format of the image stored in the memory, it can be determined whether the image type is a color, monochrome, or grayscale image.

  Next, the image storage processing unit 103 stores and stores the image on the memory in the image storage unit 103a, and stores the file name in association with the image.

  Next, the image management information processing unit 104 issues a unique image ID to the image, the image ID and the file name of the image (file name including full path information indicating the storage location (address)), Information associated with the date at the time of image input is stored as management information in the image management database (DB) 104a.

  An example of the configuration of the image management DB 104a is shown in FIG.

  Next, the luminance feature information extraction processing unit 105 extracts luminance feature information related to the luminance of the image, and registers the extracted luminance feature information and the image ID in the luminance feature information index table 105a in association with each other. Further, the color feature information extraction processing unit 106 extracts color feature information relating to the color of the image, and associates the extracted color feature information with the image ID and registers them in the color feature information index table 106a.

  An example of the configuration of the luminance feature information index table 105a and the color feature information index table 106a is shown in FIGS. 3 and 4, respectively. In these index tables, various feature information and image IDs are stored and managed in pairs, so that correspondence can be established by intermediating records in the image management DB 104a with image IDs.

  Further, the color feature information extraction processing unit 106 extracts the color feature information even when the type of the processing target image is a monochrome or gray scale image. In particular, when the comparison source image is a color image, a similar image search process using the color feature information is executed. At this time, the comparison destination image is a monochrome or grayscale image. This is because if there is no color information, the color feature information indicating this can be extremely important information.

  For example, when the comparison source image is composed of a light color, for example, pastel color, even if the comparison destination image is a monochrome or grayscale image, the color feature information of the comparison source image and the comparison destination image (search This is because there is an expectation that a high degree of similarity is calculated with the color feature information of the target image.

  Next, details of processing executed by the luminance feature information extraction processing unit 105 will be described.

<Luminance feature information extraction processing>
Here, information associating the luminance class corresponding to the average luminance of each block obtained by dividing the processing target image into a plurality of blocks and the position information of each block is extracted as luminance feature information.

  Details of this processing will be described below with reference to FIG.

  FIG. 5 is a flowchart showing details of the luminance feature information extraction processing according to the embodiment of the present invention.

  First, in step S501, it is determined whether or not the processing target image is a color image.

  This determination is performed based on the data format of the processing target image.

  For example, in the DIB format which is a general-purpose image data format used in Windows (registered trademark) which is an operating system provided by Microsoft, biBitCount which is a member of the structure BITMAPINFOHEADER in the data header is 1 in the image. The number of bits used to represent a pixel is shown.

  In particular, when the biBitCount is 24, a full color image is shown, when the biBitCount is 1, a monochrome binary image is shown, and when the biBitCount is 8, a 256 color or grayscale image is shown.

  When biBitCount is 8, it is possible to determine whether the image is a 256 color image or a gray scale image by referring to the contents of the member bmiColors of the structure BITMAPINFO in the data header. In other words, the member bmiColors indicates whether the content is a gray scale expression color palette or a color palette that is reduced to 256 colors, and based on the content, the image is displayed in 256 colors or gray. Whether the image is a scale image can be determined.

  In step S501, when the processing target image is not a color image (NO in step S501), the process proceeds to step S503. On the other hand, if the processing target image is a color image (YES in step S501), the process advances to step S502 to convert the color image into a grayscale image (8 bits: 256 gradations).

This conversion is performed using a known RGB color matrix. For example, when the YCbCr color space is used, the relationship between the luminance Y value indicating the gray scale value and the RGB value (8 bits each, 24 bits in total) is
Y = 0.29900 * R + 0.58700 * G + 0.11400 * B (1)
It is expressed. Therefore, the value of the luminance Y can be calculated by the equation (1).

  Next, in step S503, the image is divided into a plurality of blocks.

  In the present embodiment, as shown in FIG. 6, the image is divided into 9 blocks vertically and horizontally. In particular, in the present embodiment, an example is shown in which 9 × 9 = 81 blocks are divided for convenience of description, but actually 8 to 16 blocks are preferable.

  In step S504, the target block to be processed is set as the upper left block. For example, as shown in FIG. 7, the block of interest is set with reference to an order determination table in which the processing order is determined in advance. In this example, scanning is performed one right from the upper left corner, and scanning is performed obliquely from the block to the lower left. The scanning example which scans by repeating this is shown.

  In step S505, it is determined whether there is an unprocessed block of interest. If there is no unprocessed block of interest (NO in step S505), the process ends. On the other hand, if there is an unprocessed block of interest (YES in step S505), the process proceeds to step S506.

  In step S506, the average brightness of the brightness of all the pixels in the block of interest is calculated. In step S507, it is determined whether or not the calculated average luminance is equal to or higher than a luminance threshold Th_G. If the average luminance is less than the luminance threshold Th_G (NO in step S507), the process proceeds to step S508.

  In step S508, the brightness class ID corresponding to the average brightness is determined as the representative brightness of the target block with reference to the brightness class table of FIG. 8, and the brightness feature information index is associated with the target block and its position. Store in table 105a.

  Note that the luminance class table of FIG. 8 is composed of various luminance classes in which an 8-bit luminance signal is set for each predetermined luminance range, and luminance classes indicating character blocks.

  In step S509, the next block of interest is set with reference to the order determination table of FIG. Thereafter, the process returns to step S505, and the processes in steps S505 to S512 are recursively repeated until there is no unprocessed block of interest.

  Note that the reason why it is determined in step S507 whether or not the average brightness is equal to or higher than the brightness threshold Th_G is that the background of the print document image is usually white and white is expressed in a non-printed portion. This is because it may be assumed that the portion above the appropriate threshold Th_G is a plain background or a background with characters. Regarding this threshold Th_G, for example, it is plain like recycled paper, but it is necessary to take into account the low brightness when scanned.

  On the other hand, if the calculated average brightness is equal to or greater than the brightness threshold Th_G in step S507 (YES in step S507), it is verified in step S510 whether there is a character in the block of interest. In step S511, the presence / absence of characters is determined based on the verification result. If there is no character (NO in step S511), the process proceeds to step S508.

  On the other hand, if there is a character (YES in step S511), in step S512, a special luminance class ID indicating that the block of interest is a character block, in FIG. 8, the luminance class ID 17 is associated with the block of interest and its position. Stored in the luminance feature information index table 105a.

  Through the above processing, information indicating the representative luminance or character block corresponding to the average luminance for each block of the processing target image and the information in which the position information of each block is associated can be extracted as luminance feature information.

  Next, details of the process for determining whether or not there is a character in the block of interest in step S510 will be described with reference to FIG.

<Determining whether or not there is character information in the block of interest>
FIG. 9 is a flowchart showing details of the process for determining whether or not there is a character in the block of interest in step S510 according to the embodiment of the present invention.

  In step S601, a grayscale image of the block of interest is acquired from the grayscale image already acquired in step S502. In step S602, using the threshold value Th_Char, the run length distribution of pixels exceeding the threshold value Th_Char is obtained in both the horizontal direction and the vertical direction.

  The range for obtaining the run-length distribution may be a range that slightly exceeds the assumed character size. As for the threshold value Th_Char, the average luminance of the block of interest may be used. For simplicity, the threshold value Th_G may be used instead.

  Further, in order to correspond to a color ground document image of a certain density, the highest luminance pixel of the document image and its luminance are acquired in advance, and a value obtained by adding a threshold value Th_G to the value is used as a new Th_G. May be.

  In general, when a region where characters exist is statistically analyzed with a run-length distribution, the distribution has a sharp peak near the size of the character font as shown in FIG. 10, and the distribution sharply decreases near the size of the character font. A characteristic distribution is obtained.

  By using this property, it is possible to estimate the presence or absence of characters by detecting the dispersion of the peak of the run-length distribution and the slope of the steep negative distribution.

  Therefore, first, in step S603, the dispersion of the peak of the run-len length distribution in both the horizontal and vertical directions is calculated. In step S604, the steepest frequency decrease gradient is calculated in a region longer than the run length corresponding to the mode value of the run length distribution in both the horizontal and vertical directions.

  In step S605, it is determined whether or not the calculated variance is greater than or equal to a threshold and the calculated decrease gradient is greater than or equal to the threshold. If the calculated variance is equal to or greater than the threshold and the calculated decrease gradient is not equal to or greater than the threshold (NO in step S605), the process proceeds to step S606, and it is determined that there are no characters in the block of interest. On the other hand, when the calculated variance is equal to or greater than the threshold and the calculated decrease gradient is equal to or greater than the threshold (YES in step S605), the process proceeds to step S607, and it is determined that there is a character in the block of interest.

  Through the above processing, it is possible to roughly determine whether or not there is a certain amount of characters in the block of interest. Furthermore, the accumulated luminance information in the vertical and horizontal directions of the block of interest is generated, and the size of the character font is determined. If an autocorrelation function with an upper limit of several times the estimated value is calculated, if there is a character string, it becomes possible to detect a significant peak at a value corresponding to the column width of the character. The presence / absence determination accuracy can be improved.

  Next, details of processing executed by the color feature information extraction processing unit 106 will be described.

  Intuitively, the color feature information extraction may be considered as an extension of the above-described luminance feature information extraction to three dimensions. For example, in the luminance feature information extraction, the representative luminance of each block of interest is determined using the one-dimensional luminance class table of FIG. 8, but in the color feature information extraction, the RGB color space as shown in FIG. It is divided into a plurality of partial spaces (so-called color bins), and color feature information is extracted for each color bin.

  Details of this processing will be described below with reference to FIG.

  FIG. 12 is a flowchart showing details of the color feature information extraction processing according to the embodiment of the present invention.

  First, in step S1201, it is determined whether or not the processing target image is a color image. This determination is performed in the same manner as step S501 in FIG.

  In step S1201, when the processing target image is a color image (YES in step S1201), the process proceeds to step S1203. On the other hand, if the processing target image is not a color image (NO in step S1201), the process proceeds to step S1202, and the grayscale image is converted into a color image.

  In this conversion from a grayscale image to a color image, for example, the pixels having the luminance value a may be RGB values (a, a, a), and the data arrangement may be adjusted in consideration of the DIB format padding. This padding is an image data expression specification for storing color image data in the WINDOWS (registered trademark) DIB format (bitmap file (* .BMP) that exists as a file).

  In step S1203, the image is divided into a plurality of blocks. In the present embodiment, as shown in FIG. 6, the image is divided into 9 blocks in the vertical and horizontal directions. In particular, in the present embodiment, an example in which it is divided into 9 × 9 = 81 blocks for convenience of description is shown, but actually, it is preferable that there are about 15 × 15 = 225 blocks.

  In step S1204, the target block to be processed is set as the upper left block. The target block is set in the same manner as step S1204 in FIG.

  In step S1205, the presence / absence of an unprocessed block of interest is determined. If there is no unprocessed block of interest (NO in step S1205), the process ends. On the other hand, if there is an unprocessed block of interest (YES in step S1205), the process advances to step S1206.

  In step S1206, it is determined whether the block of interest is a character block. If the block of interest is not a character block (NO in step S1206), the process proceeds to step S1207.

  This determination is made with reference to the luminance class ID of the block of interest.

  In step S1207, an average color of the colors of all pixels in the block of interest is calculated. In step S1208, the calculated average color is projected onto a color bin which is a partial space created by dividing the color space of FIG. 11, and the corresponding color bin ID is determined as the representative color of the block of interest. Then, the color bin ID and the block of interest are associated with the position and stored in the color feature information index table 106a.

  In the present embodiment, as shown in FIG. 11, the density value of all the pixels of the block of interest is projected onto a color bin obtained by dividing the RGB color space into 3 × 3 × 3 = 27. For this, it is preferable to project the average color of the block of interest onto a color bin obtained by dividing the RGB color space into 6 × 6 × 6 = 216.

  In step S1209, the target block to be processed next is set with reference to the order determination table of FIG. Thereafter, the process returns to step S1205, and the processes in steps S1205 to S1210 are recursively repeated until there is no unprocessed block of interest.

  On the other hand, if the block of interest is a character block in step S1206 (YES in step S1206), in step S1210, the color bin ID indicating the character block is associated with the block of interest and the color feature information index table 106a. To remember.

  If 6 × 6 × 6 = 216 color bins are used, for example, a color bin ID 217 may be used as a color bin ID indicating a character block.

  Through the above processing, information indicating the representative color or character block corresponding to the average color of each block of the processing target image and the position information of each block can be extracted as color feature information.

[Image search processing]
In the image search process, first, the comparison source image input unit 107 inputs a comparison source image serving as a search condition. The input of the comparison source image is realized, for example, by using an input device such as a scanner or by reading an image stored in an external storage device. The input image is temporarily stored in a memory, and from the data format of the image stored in the memory, it can be determined whether the image type is a color image, a monochrome image, or a grayscale image. .

  Note that the image search processing of the present embodiment focuses only on whether or not the color information of the comparison source image is sufficient, and if the color information of the comparison source image is sufficient, the color feature information indicates the comparison target image. If the color information of the comparison source image is insufficient, a similar image search is performed to perform a similarity comparison between the luminance feature information and the luminance feature information of the comparison destination image. The case where it is performed will be described as an example.

  Therefore, first, the determination processing unit 108 determines whether the color information of the image on the memory is sufficient.

  When the color information is sufficient, the color feature information comparison processing unit 109 uses the color feature information of the comparison source image and the color feature information of the search target image (comparison target image) stored in the image storage unit 103a. Similar image search is performed by comparing the two.

  On the other hand, when the color information is insufficient, the luminance feature information comparison processing unit 110 uses the luminance characteristic information of the comparison source image to determine the luminance of the search target image (comparison target image) stored in the image storage unit 103a. Similar image retrieval is performed by comparing with feature information.

  Accordingly, the similarity comparison process can be performed by a method suitable for the type of image of the comparison source image (monochrome, grayscale image, or color image).

  The simplest method of determination by the determination processing unit 108 is to analyze the data format of the comparison source image in the same manner as described in the above image registration process, and the comparison source image is monochrome or grayscale. In the case of an image, there is no color information, so it is determined that the color information of the comparison source image is insufficient.

  On the other hand, if the comparison source image is a color image, it may be determined that the color information of the comparison source image is sufficient. However, although the data format itself is a color image, a grayscale image is included therein. There can be. Therefore, in such a case, even when the data format is a color image, it is necessary to strictly determine whether or not the color information of the comparison source image is sufficient. There is a method of using the analysis on the color of.

  The purpose of this analysis is to analyze the ratio of color information in the comparison source image and determine that the color information is sufficient when the ratio of color information in the comparison source image is equal to or greater than a predetermined threshold. If it is less than the threshold, it is determined that the color information is insufficient.

  Here, there are two types of analysis methods for analyzing the ratio of the color information in the comparison source image in the determination processing unit 108. Therefore, these two types of analysis methods will be described below.

<Color analysis method 1>
FIG. 13 is a flowchart showing a first color analysis method according to the embodiment of the present invention.

  First, in step S1301, it is determined whether or not the comparison source image is a color image. If the comparison source image is not a color image (NO in step S1301), the process proceeds to step S1302, and it is determined that the color information of the comparison source image is insufficient. On the other hand, if the comparison source image is a color image (YES in step S1301), the process advances to step S1303.

  In step S1303, it is determined whether the number of pixels of the comparison source image is equal to or greater than a threshold value. If it is less than the threshold (NO in step S1303), the process proceeds to step S1309. On the other hand, if it is equal to or greater than the threshold (YES in step S1303), the process proceeds to step S1304.

  In step S1304, a similarity ratio (reduction ratio) is calculated such that the total number of pixels is close to a predetermined value smaller than that number.

For example, if the target number of pixels is Gp, the vertical size of the comparison source image is H, and the horizontal size is W,
Similarity ratio S = Int (sqrt (H * W / Gp))
It becomes.

  However, the function Int (x) means a value that is larger than x and selects an integer value.

  Therefore, the reduced image of the comparison source image is vertical H2 = H / S [pixel] and horizontal W2 = W / S [pixel].

  Next, in step S1305, the average color of the processing target block of the pixel size of the comparison source image in the vertical and horizontal S × S blocks is calculated, and this is used as the pixel value of the corresponding pixel in the reduced image. The processing target blocks are moved so as not to overlap, and a reduced image of the comparison source image is generated by repeating this processing.

  The reason why the reduced image of the comparison source image is generated in step S1305 is that when the color histogram of the comparison source image is generated in the following step S1306, the color histogram is generated when the number of pixels of the processing target image is very large. This is because the processing load increases, so that the number of pixels of the comparison source image is set to a desired number so that the processing load is reduced. Of course, when the number of pixels of the processing target image is small, a color histogram may be generated as it is.

  Next, in step S1306, the density values of all the pixels constituting the reduced image are projected onto a color bin, which is a partial space created by dividing the color space of FIG. 11, and a color histogram for the color bin is generated.

  In step S1307, the number of pixels belonging to the color bin in the gray scale direction is counted from the histogram information for each color bin. In step S1308, the ratio of the number of pixels belonging to the color bin in the grayscale direction to the total number of pixels of the reduced image is calculated.

  In step S1312, it is determined whether the calculated ratio is equal to or greater than a predetermined threshold. If it is equal to or greater than the threshold (YES in step S1312), the process proceeds to step S1302. On the other hand, if it is less than the threshold (NO in step S1312), the process advances to step S1313 to determine that the color information of the comparison source image is sufficient.

  On the other hand, if the number of pixels of the comparison source image is less than the threshold value in step S1303, the density values of all the pixels constituting the comparison source image are created by dividing the color space of FIG. 11 in step S1309. Project to a color bin which is a subspace, and generate a color histogram for the color bin. In step S1310, the number of pixels belonging to the color bin in the gray scale direction is counted from the histogram information for each color bin. In step S1311, the ratio of the number of pixels belonging to the color bin in the grayscale direction to the total number of pixels is calculated. Thereafter, the process proceeds to step S1312.

<Color analysis method 2>
FIG. 14 is a flowchart showing a second color analysis method according to the embodiment of the present invention.

  First, in step S1401, it is determined whether or not the comparison source image is a color image. If the comparison source image is not a color image (NO in step S1401), the process proceeds to step S1402, and it is determined that the color information of the comparison source image is insufficient. On the other hand, if the comparison source image is a color image (YES in step S1401), the process advances to step S1403.

  In step S1403, an average color of all the pixels constituting the comparison source image is calculated. In step S1404, the average color is converted into a luminance component and a color difference component. In step S1405, the ratio R of the color difference component value to the luminance component value is calculated.

  Here, a known method is used as a separation method for separating a color into a luminance component and a color difference component.

For example, using the YCbCr color space, the association with a 24-bit RGB value is
Y = 0.29900 * R + 0.58700 * G + 0.11400 * B
Cb = -0.16874 * R-0.33126 * G + 0.50000 * B + 128
Cr = 0.50000 * R-0.41869 * G + (-0.08131) * B + 128 (2)
It is expressed.

According to the equation (2), the calculated average color is separated into a luminance component Yave and color difference components Cbave and Crave,
Ratio R = sqrt (Cbave * Cbave + Crave * Crave) / Yave (3)
Calculate

  In step S1406, it is determined whether the ratio R is equal to or greater than a predetermined threshold. If it is equal to or greater than the threshold (YES in step S1406), the process advances to step S1402. On the other hand, if it is less than the threshold (NO in step S1406), the process advances to step S1407 to determine that the color information of the comparison source image is sufficient.

  Next, details of processing executed by the color feature information comparison processing unit 109 will be described.

  The color feature information comparison processing unit 109 uses the above-described color feature information extraction processing in order to extract the color feature information of the comparison source target image.

  FIG. 15 is a flowchart showing details of the color feature information comparison processing according to the embodiment of the present invention.

  First, in step S1501, the color feature information of the comparison source image is extracted for each block by the processing described in the flowchart of FIG. However, if the comparison source image is already registered in the image storage unit 103a, the corresponding color feature information is read from the color feature index table 106a.

  Next, in step S1502, the presence / absence of a comparison source image and an uncompared comparison destination image (stored in the image storage unit 103a) is determined. If there is no uncompared comparison destination image (NO in step S1502), the process advances to step S1511. On the other hand, if there is an uncompared comparison destination image (YES in step S1502), the process advances to step S1503.

  In step S1503, with reference to the color feature information index table 106a, the image ID of the processing target image and the color feature information of each block are read. In step S1504, the target block in the image to be processed is set at the head. In step S1505, the similarity distance indicating the similarity between the color feature information of the comparison source image and the color feature information to be compared is reset to zero.

  In step S1506, it is determined whether there is an uncompared block of interest. If there is no uncompared block of interest (NO in step S1506), the process advances to step S1510. On the other hand, if there is an uncompared block of interest (YES in step S1506), the process advances to step S1507.

  In step S1507, the color bin ID of each block of interest is acquired from the color feature information of each of the comparison source image and the comparison destination image.

  In step S1508, referring to the color bin penalty matrix of FIG. 16, the local similarity distance of the block of interest corresponding to the acquired color bin ID is acquired, and this is accumulated in the similar distance acquired in the immediately preceding process. to add. The accumulated similarity distance is stored in the memory.

  Here, the color bin penalty matrix will be described with reference to FIG.

  FIG. 16 is a diagram showing the configuration of the color bin penalty matrix according to the embodiment of the present invention.

  The color bin penalty matrix is a matrix that manages the local similarity distance between the color bin IDs. According to FIG. 16, the similarity distance of the color bin penalty matrix is 0 for the same color bin ID, and the similarity distance increases as the difference between the color bin IDs increases, that is, as the similarity decreases. It is configured. In addition, all the diagonal positions of the same color bin ID have a similarity distance of 0 and have symmetry with respect to the boundary.

  However, in this example, the color bin ID 217 is set for the color bin ID corresponding to the character block. Since this color bin ID 217 has different properties with respect to the penalty between the other color bin IDs, a relatively large penalty, for example, between white and black is set. If it does in this way, it will become a similar comparison process which considers a difference very severely regarding a character block. That is, in the color bin penalty matrix of FIG. 16, the similar distance between the predetermined color bin ID indicating the character block and the color bin ID of the block other than the character block is the similar distance between the color bin IDs other than the predetermined color bin ID. Are managed to be sufficiently different (in this case, large).

  Conversely, if the penalty between the color bin ID corresponding to the character block and the other color bin IDs is reduced, similar comparison processing is performed in which a difference is loosely considered for the character block. Therefore, by adjusting the penalty value as necessary, it is possible to realize a similar comparison process that is more suitable for the purpose and purpose.

  Thus, in this embodiment, the similarity distance between the color bin IDs can be acquired only by referring to the color bin penalty matrix, so that the processing can be speeded up.

  In step S1509, the target block to be processed next is set.

  On the other hand, if there is no uncompared block of interest in step S1506 (NO in step S1506), the process proceeds to step S1510, and the accumulated similarity distance stored in the memory is stored in the memory in association with the image ID.

  If there is no uncompared comparison target image in step S1502 (NO in step S1502), the process proceeds to step S1511, and the image IDs are sorted in ascending order of the cumulative similarity distance corresponding to each image ID stored in the memory. Then, the sorted image ID and cumulative similarity distance pair is output as a search result.

  Next, details of processing executed by the luminance feature information comparison processing unit 110 will be described.

  The luminance feature information comparison processing unit 110 uses the above-described luminance feature information extraction processing in order to extract the luminance feature information of the comparison source target image.

  FIG. 17 is a flowchart showing details of the luminance feature information comparison processing according to the embodiment of the present invention.

  First, in step S1701, the luminance class ID is extracted as the luminance feature information of the comparison source image by the processing described in the flowchart of FIG. However, if the comparison source image is an image that has already been registered in the image storage unit 103a, the corresponding luminance feature information is read from the luminance feature index table 105a.

  Next, in step S1702, it is determined whether there is a comparison source image and an uncompared comparison destination image (stored in the image storage unit 103a). If there is no uncompared comparison destination image (NO in step S1702), the process advances to step S1711. On the other hand, if there is an uncompared comparison destination image (YES in step S1702), the process advances to step S1703.

  In step S1703, the luminance feature information index table 105a is referred to, and the image ID of the processing target image and the luminance feature information of each block are read out. In step S1704, the target block in the image to be processed is set at the head. In step S1705, the similarity distance indicating the similarity between the luminance feature information of the comparison source image and the luminance feature information to be compared is reset to zero.

  In step S1706, it is determined whether there is an uncompared block of interest. If there is no uncompared block of interest (NO in step S1706), the process advances to step S1710. On the other hand, if there is an uncompared block of interest (YES in step S1706), the process proceeds to step S1707.

  In step S1707, the brightness class ID related to the brightness of the block of interest is acquired from the brightness feature information of each of the comparison source image and the comparison destination image.

  In step S1708, referring to the luminance class ID penalty matrix in FIG. 18, the local similarity distance of the target block corresponding to the acquired luminance class ID is acquired, and this is used as the similar distance acquired in the immediately preceding process. Cumulative addition. The accumulated similarity distance is stored in the memory.

  Here, the luminance class ID penalty matrix will be described with reference to FIG.

  FIG. 18 is a diagram showing a configuration of a luminance class ID penalty matrix according to the embodiment of the present invention.

  The luminance class ID penalty matrix is a matrix that manages the local similarity distance between luminance class IDs. According to FIG. 18, the luminance class ID penalty matrix has a similarity distance of 0 for the same luminance class ID, and the similarity distance increases as the difference between the luminance class IDs increases, that is, as the similarity decreases. It is configured. Further, all the diagonal positions of the same luminance class ID have a similarity distance of 0 and have symmetry with respect to the boundary.

  However, the luminance class ID corresponding to the character block, in this example, the luminance class ID 17 is set. This luminance class ID 17 has a different characteristic with respect to the penalty between other luminance class IDs, and therefore, a relatively large penalty, for example, between white and black is set. If it does in this way, it will become a similar comparison process which considers a difference very severely regarding a character block. That is, in the luminance class ID penalty matrix of FIG. 18, the similarity distance between the predetermined luminance class ID indicating a character block and the luminance class ID of a block other than the character block is similar to the luminance class IDs other than the predetermined luminance class ID. It is managed to be sufficiently different (in this case, larger) than the distance.

  Conversely, if the penalty between the luminance class ID corresponding to the character block and the other luminance class IDs is reduced, similar comparison processing is performed in which the difference is loosely considered for the character block. Therefore, by adjusting the penalty value as necessary, it is possible to realize a similar comparison process that is more suitable for the purpose and purpose.

  As described above, in this embodiment, the similarity distance between the luminance class IDs can be acquired only by referring to the luminance class ID penalty matrix, so that the processing speed can be increased.

  In step S1709, the target block to be processed next is set.

  On the other hand, if there is no uncompared block of interest in step S1706 (NO in step S1706), the process proceeds to step S1710, and the accumulated similarity distance stored in the memory is stored in the memory in association with the image ID.

  If there is no uncompared image in step S1702 (NO in step S1702), the process proceeds to step S1711, and the image IDs are sorted in ascending order of the cumulative similarity distance corresponding to each image ID stored in the memory. The sorted image ID and cumulative similarity distance pair is output as a search result.

  By executing the image registration process and the image search process described above, an image feature amount suitable for the type of image can be obtained regardless of whether the type of the comparison target image is monochrome, grayscale, or a color image. Use to register images or search for images.

  Next, search result display processing executed by the search result display unit processing unit 111 will be described.

[Search result display processing]
In the search result display process, the image management DB 104a is referred to based on information in which the pairs of image IDs and similarities are sorted in descending order of similarity, that is, in ascending order of similarity distances, and images are stored from corresponding records. The previous address information (file full path information) is acquired, and the corresponding image is read from the image storage unit 103a and displayed.

  The search result display process will be described below with reference to FIG.

  FIG. 19 is a flowchart showing search result display processing according to the embodiment of the present invention.

  In step S1801, the number of search hits M, which is the number of search results obtained by the image search process, corresponds to the descending order of similarity between the comparison source image and the comparison destination image obtained as the search result, that is, the ascending order of the similarity distance. A pair of the image ID and its similarity is acquired.

  In step S1802, the number N of images that can be displayed on the display screen is acquired.

  This number N is determined by the size of the reduced image of the image to be displayed as the search result and the size of the search result display area on the display screen.

  In step S1803, the smaller of the search hit number M and the number N is set as the image display number R.

  In step S1804, a variable K indicating the order of images to be displayed as a search result (read image order) is initialized to 1. In step S1805, it is determined whether K ≦ R. If K ≦ R is not satisfied (NO in step S1805), the process ends. On the other hand, if K ≦ R (YES in step S1806), the process proceeds to step S1806.

  Here, the processing from step S1806 to step S1808 is processing for displaying an image as a similar image search result on the display screen.

  Here, a display example of the similar image search result will be described with reference to FIG.

  FIG. 20 is a diagram showing a display example of similar image search results according to the embodiment of the present invention.

  As shown in FIG. 20, in the search result display area on the display screen, the reduced images of the search results are displayed side by side in order of increasing similarity distance from the upper left to the right next to the upper right. Then, after moving to the left end of the next line and then displaying the next right next to the right end in sequence, the display is similarly performed until the lower right end at the end of the search result display area. A corresponding similar distance is also displayed below the reduced image to be displayed.

  In step S1806, the display position of the display target image is determined based on the read image rank K. In step S1807, image address information (full path information) corresponding to the image ID of the read image rank K is acquired from the image management DB 104a. In step S1808, the image indicated by the address information is read, and the reduced image is displayed at the determined display position. Thereafter, the variable K is incremented by 1, and the process proceeds to step S1805.

  Note that the image search process of the present embodiment focuses only on whether the color information of the comparison source image is sufficient, and if the color information of the comparison source image is sufficient, the color feature information of the comparison target image is used. Similarity search with color feature information is performed, and when the color information of the comparison source image is insufficient, an example of a similar image search is performed in which the similarity comparison between the luminance feature information and the luminance feature information of the comparison destination image is performed. However, it is not limited to this.

  For example, considering whether or not the color information of the comparison target image is sufficient, if the color information of both images to be compared is sufficient, a similarity comparison is performed using both color feature information, and either one or When both pieces of color information are insufficient, a similar image search may be performed in which similarity comparison is performed using both luminance feature information.

  In the above embodiment, it is configured to determine whether or not a character image exists in the target block based on the average luminance in the target block. However, the present invention is not limited to this. For example, whether or not a character image exists in the target block is determined based on the mode luminance in the block of interest, or the average luminance or mode luminance of the color neighborhood range of the pixel with the highest luminance in the block of interest. It is good also as composition to do.

  In the case of this configuration, before performing the luminance feature information extraction processing, preprocessing related to the luminance histogram is performed on the block of interest, normalization is performed from luminance values 0 to 255, and then the luminance feature information extraction processing is executed. Further simplification of processing can be achieved.

  Thus, for example, a plain background such as recycled paper that is pre-colored and has no ink or toner on it is normalized to a luminance value of 255, and even if this embodiment is applied, it is accurately plain. It becomes possible to extract characters printed on the background.

  The image registration processing is performed in the order of image input processing, image storage processing, image management information processing, luminance feature information extraction processing, and color feature information extraction processing. Does not matter.

  On the other hand, in the image search processing, if the comparison source image is an image that has already been registered in the image storage unit 103a, the color feature information and luminance feature information have already been registered. Processing and luminance feature information extraction processing are not required.

  In addition, the image management DB 104a, the luminance feature information index table 105a, and the color feature information index table 106a are configured separately, and the configuration example in which the image ID is used as an intermediary to correspond to each other's information has been shown. May be configured integrally.

  Further, the setting of the target block to be processed is not limited to the scanning order shown in the order determination table of FIG. 7, but other scanning orders such as horizontal scanning and vertical scanning or zigzag scanning, that is, image Any scanning order may be used as long as each block can be set as a target block.

  As described above, according to the present embodiment, an area like a method for realizing image search by obtaining features of separated objects using a technique for separating a document image into a character area and an image area in advance. A document image comparison process in which a rough character detection result is added to a primitive image feature amount is realized without performing a heavy load process that may cause a search omission due to failure to extract.

  This makes it possible to search for a target image with higher accuracy, especially for an image in which an image is pasted on a white background and characters are present in the margin, such as a document image.

  In other words, even if there is a character on the white background, the average color or the most frequent color is not affected, and the divided block with the character cannot be distinguished from the white background, making the search uncomfortable for the user. It is possible to prevent the result.

  Although the embodiments have been described in detail above, the present invention can take an embodiment as, for example, a system, an apparatus, a method, a program, or a storage medium, and specifically includes a plurality of devices. The present invention may be applied to a system that is configured, or may be applied to an apparatus that includes a single device.

  In the present invention, a software program (in the embodiment, a program corresponding to the flowchart shown in the figure) that realizes the functions of the above-described embodiment is directly or remotely supplied to the system or apparatus, and the computer of the system or apparatus Is also achieved by reading and executing the supplied program code.

  Accordingly, since the functions of the present invention are implemented by computer, the program code installed in the computer also implements the present invention. In other words, the present invention includes a computer program itself for realizing the functional processing of the present invention.

  In that case, as long as it has the function of a program, it may be in the form of object code, a program executed by an interpreter, script data supplied to the OS, or the like.

  As a recording medium for supplying the program, for example, floppy (registered trademark) disk, hard disk, optical disk, magneto-optical disk, MO, CD-ROM, CD-R, CD-RW, magnetic tape, nonvolatile memory card ROM, DVD (DVD-ROM, DVD-R) and the like.

  As another program supply method, a client computer browser is used to connect to an Internet homepage, and the computer program of the present invention itself or a compressed file including an automatic installation function is downloaded from the homepage to a recording medium such as a hard disk. Can also be supplied. It can also be realized by dividing the program code constituting the program of the present invention into a plurality of files and downloading each file from a different homepage. That is, a WWW server that allows a plurality of users to download a program file for realizing the functional processing of the present invention on a computer is also included in the present invention.

  In addition, the program of the present invention is encrypted, stored in a storage medium such as a CD-ROM, distributed to users, and key information for decryption is downloaded from a homepage via the Internet to users who have cleared predetermined conditions. It is also possible to execute the encrypted program by using the key information and install the program on a computer.

  In addition to the functions of the above-described embodiments being realized by the computer executing the read program, the OS running on the computer based on an instruction of the program is a part of the actual processing. Alternatively, the functions of the above-described embodiment can be realized by performing all of them and performing the processing.

  Furthermore, after the program read from the recording medium is written in a memory provided in a function expansion board inserted into the computer or a function expansion unit connected to the computer, the function expansion board or The CPU or the like provided in the function expansion unit performs part or all of the actual processing, and the functions of the above-described embodiments are realized by the processing.

1 is a block diagram of an image processing apparatus according to an embodiment of the present invention. It is a figure which shows an example of a structure of image management DB of embodiment of this invention. It is a figure which shows an example of a structure of the brightness | luminance characteristic information table of embodiment of this invention. It is a figure which shows an example of a structure of the color characteristic information table of embodiment of this invention. It is a flowchart which shows the detail of the brightness | luminance feature information extraction process of embodiment of this invention. It is a figure which shows an example of the image block division | segmentation of embodiment of this invention. It is a figure which shows an example of the order determination table of embodiment of this invention. It is a figure which shows an example of the brightness class table of embodiment of this invention. It is a flowchart which shows the detail of the determination process of the presence or absence of the character in the attention block of step S510 of embodiment of this invention. It is a figure which shows an example of the run length distribution of embodiment of this invention. It is a figure which shows an example of a structure of the color bin on the color space of embodiment of this invention. It is a flowchart which shows the detail of the color feature information extraction process of embodiment of this invention. It is a flowchart which shows the 1st color analysis method of embodiment of this invention. It is a flowchart which shows the 2nd color analysis method of embodiment of this invention. It is a flowchart which shows the detail of the color characteristic information comparison process of embodiment of this invention. It is a figure which shows an example of a structure of the color bin penalty matrix of embodiment of this invention. It is a flowchart which shows the detail of the brightness | luminance feature information comparison process of embodiment of this invention. It is a figure which shows an example of a structure of the brightness class ID penalty matrix of embodiment of this invention. It is a flowchart which shows the search result display process of embodiment of this invention. It is a figure which shows the example of a display of the similar image search result of embodiment of this invention.

Explanation of symbols

DESCRIPTION OF SYMBOLS 101 User interface part 102 Image input process part 103 Image storage process part 103a Image storage part 104 Image management information processing part 104a Image management DB
105 Luminance Feature Information Extraction Processing Unit 105a Luminance Feature Information Index Table 106 Color Feature Information Extraction Processing Unit 106a Color Feature Information Index Table 107 Comparison Source Image Input Unit Processing Unit 108 Judgment Processing Unit 109 Color Feature Information Comparison Processing Unit 110 Luminance Feature Information Comparison Processing unit 111 Search result display processing unit

Claims (11)

An image processing apparatus that performs a similar image search based on a similarity between a comparison source image serving as a search condition and a comparison destination image,
Dividing means for dividing the comparison source image into a plurality of blocks;
Determination means for determining whether a character image exists in the block image based on image feature information of the block image divided by the dividing means;
As a result of the determination by the determination means, when a character image exists in either the block image or the block image corresponding to the comparison target image, a predetermined similarity is acquired as the similarity between the two, while the block image And when there is no character image in either of the block images of the comparison target image corresponding thereto, acquisition means for acquiring the similarity based on the image feature information of both,
Output means for outputting an image as a search result on the basis of a cumulative similarity obtained by cumulatively adding similarities to the respective block images divided by the dividing means obtained by the obtaining means. An image processing apparatus. The comparison unit includes a management table that manages similarity according to image feature information between the comparison source image and the comparison destination image,
The management table includes the comparison source as a predetermined similarity based on image feature information between the comparison source image and the comparison destination image when a character image exists in either the comparison source image or the comparison destination image. 2. The image according to claim 1, wherein a fixed value that is sufficiently different from a similarity based on image feature information between the comparison source image and the comparison destination image when no character image exists in the image is managed. Processing equipment. The determination unit determines whether or not an image feature amount indicated by the image feature information of the block image is equal to or greater than a predetermined value. If the image feature amount is equal to or greater than a predetermined value, a character image is included in the processing target image. The image processing apparatus according to claim 1, wherein the image processing apparatus is determined to exist. The image processing apparatus according to claim 3, wherein the image feature information is luminance feature information related to the luminance of the block image. The image processing apparatus according to claim 4, wherein the luminance feature information is an average luminance or a mode luminance of the block image. The image processing apparatus according to claim 4, wherein the luminance feature information is an average luminance or a mode luminance in a color vicinity range of a pixel having the highest luminance in the block image. An image processing apparatus that performs image registration processing of an image to be searched for similar image search,
An input means for inputting an image;
Dividing means for dividing the image into a plurality of blocks;
Determination means for determining whether or not a character image exists in each block image based on image feature information of each block image divided by the dividing means;
If a character image exists in the block image as a result of the determination by the determining means, information indicating the character image is stored as image feature information. On the other hand, if a character image does not exist in the block image, the image feature of the block image is stored. An image processing apparatus comprising: storage means for storing information. A control method of an image processing apparatus that performs a similar image search based on a similarity between a comparison source image serving as a search condition and a comparison target image,
A dividing step of dividing the comparison source image into a plurality of blocks;
A determination step of determining whether or not a character image exists in the block image based on image feature information of the block image divided in the division step;
As a result of the determination in the determination step, when a character image exists in either the block image or the block image of the comparison target image corresponding thereto, a predetermined similarity is obtained as the similarity between the two, while the block And when there is no character image in either of the block images of the comparison target image corresponding thereto, an acquisition step of acquiring the similarity based on the image feature information of both,
An output step of outputting an image as a search result based on a cumulative similarity obtained by cumulatively adding the similarity to each block image divided in the division step acquired by the acquisition step. For controlling an image processing apparatus. A control method of an image processing apparatus that performs image registration processing of an image to be searched for similar image search,
An input process for inputting an image;
A dividing step of dividing the image into a plurality of blocks;
A determination step of determining whether or not a character image exists in each block image based on image feature information of each block image divided in the division step;
As a result of the determination in the determination step, if a character image exists in the block image, information indicating it is stored as image feature information. On the other hand, if a character image does not exist in the block image, the image feature of the block image And a storage step of storing information in a storage unit. A program that realizes control of an image processing apparatus that performs a similar image search based on a similarity between a comparison source image that is a search condition and a comparison target image,
A program code of a dividing step of dividing the comparison source image into a plurality of blocks;
Based on the image feature information of the block image divided in the division step, a program code of a determination step for determining whether or not a character image exists in the block image;
As a result of the determination in the determination step, when a character image exists in either the block image or the block image of the comparison target image corresponding thereto, a predetermined similarity is obtained as the similarity between the two, while the block And when there is no character image in either of the block images of the comparison target image corresponding thereto, the program code of the obtaining step for obtaining the similarity based on the image feature information of both,
A program code of an output step for outputting an image as a search result based on a cumulative similarity obtained by cumulatively adding the similarity to each block image divided in the division step acquired by the acquisition step. A program characterized by A program that realizes control of an image processing apparatus that performs image registration processing of an image to be searched for similar image search,
An input process for inputting an image;
A dividing step of dividing the image into a plurality of blocks;
A determination step of determining whether or not a character image exists in each block image based on image feature information of each block image divided in the division step;
As a result of the determination in the determination step, if a character image exists in the block image, information indicating it is stored as image feature information. On the other hand, if a character image does not exist in the block image, the image feature of the block image A storage step of storing information in a storage unit.

Images