JP2011018311A - Device and program for retrieving image, and recording medium - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide an image retrieval device, an image retrieval program and a recording medium, allowing speed-up of processing speed related to retrieval processing, and allowing improvement of retrieval accuracy of document image data.SOLUTION: In a step A1, a preprocess part 130 performs binarization processing as preprocessing to input image data. In step A2, based on the binarized image data, a characteristic extraction part 131 extracts a characteristic amount of the binarized image data. In step A3, a retrieval part 132 compares a characteristic amount of registration image data and a characteristic amount of the binarized data, and retrieves image data having high similarity to the input image data from the registration image data. In step A4, a retrieval result is output to a display part 14. Here, the characteristic amount is arrangement wherein a ratio fn of word lengths of two adjacent words is arranged along arrangement order of the words.

Description

  The present invention relates to an image search device, an image search program, and a recording medium that search for specific image data from registered image data.

  In image forming apparatuses such as copiers, facsimile machines, printers, and multifunction machines having a plurality of these functions, image data such as input original images is stored in a large-capacity storage device, which is input once and registered. Some image data are provided with a function that can be read and re-output at any time.

  Although the function that can be re-output is convenient, it becomes difficult to find the data that you want to re-output if the amount of registered data increases, so image search technology that searches for desired image data from multiple image data is important It becomes.

  When searching for image data, it is necessary to calculate the similarity by comparing the registered image data with the input image data. The registered image data includes document image data and non-image data. Document image data (photos, figures, illustration images, etc.).

  Since the document image data is composed of character images, the character images are compared with each other, and it is difficult to determine the similarity compared to the non-document image data.

  The document image search device described in Patent Document 1 recognizes punctuation marks from document image data or search document data input by an input unit, measures the number of characters between punctuation marks, registers the measured number of characters as an index, and measures An index having the same number of characters as the number of characters between the punctuation marks of the search document data is searched.

JP 2008-152502 A

  The document image search apparatus described in Patent Document 1 searches using the number of characters between punctuation marks as an index, and such an index represents relatively large layout information. Even if the number of characters between punctuation marks is the same, there can be document image data that the characters themselves are different. Therefore, when judging by the number of characters between punctuation marks, the search accuracy cannot be sufficiently increased.

  Punctuation marks are images that are smaller than characters, and there is a high possibility that misrecognition will occur when recognizing punctuation marks from registered document image data and input image data, and misrecognition occurs when recognizing punctuation marks. In this case, the accuracy of the registered index is lacking.

  To improve the punctuation recognition accuracy in order to increase the search accuracy, it is necessary to increase the reading resolution, resulting in an increase in the amount of document image data and an increase in storage capacity. This causes a decrease in processing speed.

  An object of the present invention is to provide an image search device, an image search program, and a recording medium that can increase the processing speed of search processing and improve the search accuracy of document image data.

The present invention is an image search device for searching document image data similar to input document image data from pre-registered document image data,
A word contained in the input document image data is detected and divided into words, and for each two adjacent words, a ratio of the word lengths of the two words is calculated, and the calculated ratio of the word lengths is arranged as a word sequence. A feature amount extraction unit that extracts an array of word length ratios arranged in order as feature amounts of input document image data;
A registered image storage unit that stores the registered document image data and the feature amount of the registered document image data in association with each other;
Based on the feature amount of the registered document image data and the feature amount of the input document image data generated by the feature amount extraction unit, the input is performed from the registered document image data. A search unit for searching for image data similar to the document image data,
An image search apparatus comprising: a display unit configured to display document image data similar to the input document image data among the registered document image data based on a search result by a search unit.

  In the invention, it is preferable that the feature amount extraction unit calculates a ratio of the word lengths using the number of pixels indicating the length of a region constituting the word as the word length.

In the present invention, the registered image storage unit stores the word length ratio array in association with the positional information of the word length ratio array in the registered document image data,
The search unit detects a portion in the searched document image data that matches the word length ratio array of the input document image data,
The display unit displays the detected matched portion so as to be distinguishable from other portions.

The present invention is also an image search device for searching document image data similar to input document image data from previously registered document image data,
Characters included in the input document image data are detected, character division is performed, a circumscribed rectangle circumscribing one character is detected, and a pixel density which is a ratio of pixels constituting the character in the circumscribed rectangle is calculated. A feature amount extraction unit that extracts an array of pixel densities obtained by arranging the calculated pixel densities along the word arrangement order as feature amounts of input document image data;
A registered image storage unit that stores the registered document image data and the feature amount of the registered document image data in association with each other;
Based on the feature amount of the registered document image data and the feature amount of the input document image data generated by the feature amount extraction unit, the input is performed from the registered document image data. A search unit for searching for image data similar to the document image data,
An image search apparatus comprising: a display unit configured to display document image data similar to the input document image data among the registered document image data based on a search result by a search unit.

  The present invention is also an image search program for causing a computer to function as the above-described image search apparatus.

  The present invention is also a computer-readable recording medium on which an image search program for causing a computer to function as the above-described image search apparatus is recorded.

  According to the present invention, the feature amount extraction unit detects words included in the input document image data, performs word division, and calculates the ratio of the word lengths of the two words for each two adjacent words. Then, an array of word length ratios obtained by arranging the calculated word length ratios along the word arrangement order is extracted as a feature amount of the input document image data. The registered image storage unit stores the registered document image data in association with the feature amount of the registered document image data, and the search unit stores the feature amount of the registered document image data. And image data similar to the input document image data from the registered document image data based on the feature amount of the input document image data generated by the feature amount extraction unit. Search for.

  The display unit displays document image data similar to the input document image data among the registered document image data based on the search result by the search unit.

  By using the ratio of the word lengths of two adjacent words as a feature amount, the same feature amount can be obtained even in different sentences compared to the case of searching by using the number of characters between punctuation marks as in the prior art. Since the possibility is low, the search accuracy of the document image data can be improved.

  Furthermore, since the word length is not erroneously detected even if the document image data is read at a relatively low resolution, the document image data with a low resolution can be used, the processing speed related to the search process is increased, and the document image data The storage capacity for storing can also be reduced.

  Further, since the ratio of word lengths does not change even when the image is enlarged or reduced, the search accuracy does not change depending on the magnification of the image. Therefore, it is also effective for a search process in which so-called N-up image data is searched for by reducing an image of one page to 1/2 or 1/4 and using a plurality of pages as one image data.

  According to the present invention, the feature amount extraction unit calculates the ratio of the word lengths using the number of pixels indicating the length of the area constituting the word as the word length.

  By using the number of pixels, the word length can be easily detected, and the ratio of word lengths can also be easily calculated.

  According to the invention, the registered image storage unit stores the word length ratio array in association with the positional information of the word length ratio array in the registered document image data. The search unit detects a portion in the searched document image data that matches the word length ratio array of the input document image data, and the display unit detects the detected match portion as another Display distinguishable from parts.

  Thereby, document image data including a specific sentence can be searched from the registered document image data, and a sentence content can be searched.

  According to the present invention, the feature amount extraction unit detects characters included in the input document image data, performs character division, detects a circumscribed rectangle circumscribing one character, and pixels constituting the character are the circumscribed pixels. A pixel density, which is a ratio of occupying the rectangle, is calculated, and an array of pixel densities obtained by arranging the calculated pixel densities along the word arrangement order is extracted as a feature amount of the input document image data. The registered image storage unit stores the registered document image data in association with the feature amount of the registered document image data, and the search unit stores the feature amount of the registered document image data. And image data similar to the input document image data from the registered document image data based on the feature amount of the input document image data generated by the feature amount extraction unit. Search for.

  The display unit displays document image data similar to the input document image data among the registered document image data based on the search result by the search unit.

  By using the pixel density of characters as a feature amount, it is unlikely that the same feature amount will be obtained even in different sentences compared to the case of searching with the number of characters between punctuation marks as in the prior art. The search accuracy of document image data can be improved.

  Further, since the pixel density is not erroneously detected even when the document image data is read at a relatively low resolution, the low-resolution document image data can be used, the processing speed related to the search process is increased, and the document image data The storage capacity for storing can also be reduced.

  In addition, according to the present invention, the computer can be supplied as an image search program for causing a computer to function as the above-described image search device, and can be supplied as a computer-readable recording medium on which the image search program is recorded.

2 is a block diagram showing a mechanical configuration of the image search device 10. FIG. 3 is a block diagram showing a functional configuration of the image search device 10. FIG. 3 is a block diagram showing a functional configuration of an image search unit 13. FIG. 5 is a flowchart showing search processing by an image search unit 13; It is a flowchart which shows the pre-processing of step A1 by the pre-processing part 130. It is a flowchart which shows the feature extraction process of step A2 by the feature extraction part 131. It is a schematic diagram which shows word ratio and ratio of two adjacent word lengths. It is a schematic diagram for demonstrating pixel density.

Hereinafter, preferred embodiments of the present invention will be described in detail with reference to the drawings.
FIG. 1 is a block diagram illustrating a mechanical configuration of the image search apparatus 10. The image search device 10 includes a processor 4 and an external storage device 5 that stores software for the processor 4 to perform actual processing.

  The processor 4 extracts the feature amount of the input image data (hereinafter referred to as “input image data”), and compares it with a plurality of pre-registered image data (hereinafter referred to as “registered image data”). An image search process for searching and displaying registered image data similar to the input image data is actually performed. Actual processing in the processor 4 is executed by software stored in the external storage device 5. The processor 4 is composed of, for example, a normal computer main body.

  The external storage device 5 can be composed of, for example, a hard disk that can be accessed at high speed. The external storage device 5 may be configured to use a large-capacity device such as an optical disk in order to hold a large amount of registered image data. Temporary data created at the stage of each processing step during the search process may be stored in the external storage device 5 or may be stored in a semiconductor memory built in the processor 4.

  A keyboard 1 and a display device 3 are connected to the image search device 10. The keyboard 1 is used for inputting instructions for executing various software.

  The display device 3 displays an image based on input image data and registered image data, displays a search result, and the like.

  An image scanner 2 is further connected to the image search device 10. The image scanner 2 is used to read a document on which an image is printed and to capture input image data and registered image data.

  The input image data and the registered image data can be acquired by data communication from other devices on the network via the communication I / F (interface) 6 in addition to the input from the image scanner 2. The communication I / F 6 is realized by a LAN card for connecting to a LAN (Local Area Network), a modem card for connecting to a public switched telephone network, and performing data communication.

  FIG. 2 is a block diagram showing a functional configuration of the image search apparatus 10. The image search apparatus 10 includes an input unit 12, an image search unit 13, a display unit 14, and a registered image storage unit 15.

  The input unit 12 inputs input image data and registered image data. In the hardware configuration shown in FIG. 1, the image scanner 2, the communication I / F 6, and the like functionally correspond to the input unit 12. The registered image data is image data input in advance before the input image data is input, and is stored in the registered image storage unit 15.

  FIG. 3 is a block diagram illustrating a functional configuration of the image search unit 13. The image search unit 13 includes a preprocessing unit 130, a feature extraction unit 131, and a search unit 132.

  The image search unit 13 extracts a feature amount from the input image data input by the input unit 12, and searches for an image by comparing it with the feature amount extracted in advance for the registered image data.

  FIG. 4 is a flowchart showing search processing by the image search unit 13. In step A1, the preprocessing unit 130 performs binarization processing as preprocessing for the input image data. In step A2, the feature extraction unit 131 extracts the feature amount of the binarized image data based on the binarized image data. In step A3, the search unit 132 compares the feature amount of the registered image data with the feature amount of the binarized data, and searches the registered image data for image data having a high similarity to the input image data. In step A4, the search result is output to the display unit 14.

  Hereinafter, each step will be described in detail. The preprocessing of step A1 by the preprocessing unit 130 is shown in the flowchart of FIG. 5, for example.

  When the image data is input, it is determined in step B1 whether the input image data is color image data. If it is color image data, the process proceeds to step B2, grayed based on the lightness component, converted to grayscale image data, and the process proceeds to step B3. If it is not a color image, the process proceeds to step B3 and it is determined whether or not the image is grayscale image data. If it is grayscale image data, the process proceeds to step B4, binarized using a predetermined threshold, the grayscale image data is converted to binary image data, binary image data is output in step B5, and the process is terminated. If the image data is not grayscale image data, that is, binary image data, the binary image data is output in step B5, and the process ends.

  The binary image data is so-called monochrome image data in which the pixel value of each pixel constituting the image data is 0 or 1 (white pixel or black pixel), and the density (density) of each pixel of the gray image data. ) Is thresholded to classify all pixels into black pixels and white pixels.

  In document image data, since the background (background) is generally white and the character portion is black, it can be said that the pixels classified into black pixels by the binarization process are the pixels constituting the character.

  The feature extraction processing in step A2 by the feature extraction unit 131 is shown in the flowchart of FIG. 6, for example.

  When the binary image data binarized by the preprocessing unit 130 is input in step C1, all the coupling elements in the binary image data are detected in step C2.

  A coupling element is a pixel group in which connected pixels of the same color are gathered. Whether a black pixel coupling element or a white pixel coupling element is detected depends on whether the background of the input image data is a black pixel or a white pixel. As described above, generally, the background is often a white pixel, and a character image is drawn with a black pixel. Therefore, in the present embodiment, the description will be made assuming that a black pixel coupling element is detected. When the background is a black pixel, the character image is a white character drawn with white pixels, and in this case, a combination element of white pixels may be detected.

  Whether the background is a black pixel or a white pixel can be determined by a known background determination process. For example, if the overall black pixel ratio is smaller than a predetermined ratio, the background is determined to be a white pixel. If the ratio is larger than the predetermined ratio, the background is determined to be a black pixel.

  The binding element can be detected by a known detection method. For example, for one line, a continuous portion of black pixels adjacent to each other in the line (black run) is detected, and the run length of the black run and the coordinates of the black pixels at both ends of the black run are stored for each line. Keep it. The coordinates are determined in advance, for example, with the direction parallel to the line as the x-axis and the direction orthogonal to the line as the y-axis.

  For a black run on the upper and lower lines that sandwich one target line in the y direction, if the x coordinates of the black pixels at both ends are within the x coordinate range of the black pixel coordinates at both ends in each black run of the target line, The black run and the black run whose end pixel is a black pixel whose x coordinate is within the range can be regarded as being connected in the y direction. In this way, with respect to all the image data while sequentially shifting the line of interest, the x-direction connection portion and the y-direction connection portion are detected, and the black pixel connection element is detected.

  In step C3, word division is performed based on the detected coupling element. In order to divide the word, first, a text line (character string) is extracted based on the detected combination element. A text line is extracted as a pre-stage for detecting the position of the word in the input document image. For example, a run-length smoothing algorithm (RLSA) can be used to extract the text line. Here, the detected coupling element becomes a foreground separator in RLSA, and is reconstructed as a text line from the coupling element.

  When the text line is extracted, the text line is divided into words. In the present embodiment, a character is composed of black pixels, and a portion other than the character is a white pixel. Therefore, when attention is paid to a horizontal pixel row, the length of a white pixel region (segment) is extracted. Then, an average value of the lengths of these extracted white pixel areas is calculated, and a white pixel area smaller than the average value is set as an area between characters.

  Since the black pixel areas on both sides of the white pixel area between the characters are black pixel areas constituting a character included in one same word, a set of these white pixel areas and black pixel areas becomes one word.

  Word division can be performed by applying such processing to all text lines.

  In step C4, the ratio of word lengths between two adjacent words is calculated to extract the feature amount of the image data.

  The word length of each word divided by word division is represented by the number of pixels in the x direction along the line. Pixels located at both ends of one word are detected, and the number of pixels arranged between the pixels (including pixels at both ends) is set as the word length.

  FIG. 7 is a schematic diagram showing a word length and a ratio of two adjacent word lengths. The example shown in FIG. 7 shows a part of the document image data extracted, and shows a portion described as “Based on the result” in the document image data.

  By word division, the pixel group 20 corresponding to “Based”, the pixel group 21 corresponding to “on”, the pixel group 22 corresponding to “the”, and the pixel group 23 corresponding to “result” are divided.

  D1 is detected as the word length of the pixel group 20, d2 as the word length of the pixel group 21, d3 as the word length of the pixel group 22, and d4 as the word length of the pixel group 23. At this time, the ratio fn of the word lengths between two adjacent words is calculated as fn = dn / dn + 1. n is the number of words included in one line or the total number of words included in one page. In the example shown in FIG. 7, the word length ratio fn is calculated as f1 = d1 / d2, f2 = d2 / d3, and f3 = d3 / d4.

  The array of the calculated word length ratios fn is obtained by arranging the word length ratios along the word arrangement order. In the example of FIG. 7, the array of f1, f2, f3, and f4 is extracted as a feature quantity. Is done.

  When extracting an array of word length ratio fn for each line, an array of word length ratios fn is extracted for all lines included in one image data, and the feature amount of one image data is determined by all the arrays. Composed.

  In step C5, the extracted feature amount is output to the search unit 132. The image search process in step A3 by the search unit 132 is performed as follows, for example.

  In the search unit 132, the feature amount for searching the input image data obtained as described above is compared with the feature amount of the registered image data extracted in advance, and the input image data and the registered image data are compared based on the comparison result. Find the similarity of. The registered image data having the highest similarity is selected from the registered image data, and is used as a search result.

  As a search result, not only registered image data with the highest similarity but also a predetermined number of registered image data from the higher similarity may be selected and used as a search result.

  When the search result is output by the search unit 132, the display unit 14 displays an image obtained by visualizing the registered image data selected as the search result.

  The registered image data is extracted in the registered image storage unit 15 by extracting the feature amount as described above at the time of registration and associating it with the feature amount.

  Since the feature quantity of the present invention is an array composed of a plurality of word length ratios fn, the similarity is obtained even if the feature quantity associated with the registered image data does not completely match the feature quantity of the input image data. Can be sought.

  For example, when the input image data is a part of the registered image data, the feature amount of the input image data does not completely match the feature amount of the registered image data. It is included as a part of the feature amount of the image data.

  In addition, when a part of the input image data and a part of the registered image data overlap, the feature amount of the input image data does not completely match the feature amount of the registered image data, and the input image data A part of the feature amount overlaps with a part of the feature amount of the registered image data.

  Therefore, it is possible to search for similar image data from registered image data even if the feature values do not completely match, and how the features of the input image data match the feature values of the registered image data. By doing so, it is possible to detect a matching portion between the input image data and the registered image data.

  If the array of word length ratios fn and the position information (pixel coordinates) of each image data are stored in association with each other for the input image data and registered image data, an array of matching word length ratios fn is detected. Thus, a matching portion between the input image data and the registered image data can be detected based on the position information corresponding to the matched array.

  Assuming that the input image data is a part of the text content, it is possible to search for the content by detecting a matching portion from the registered image data and displaying the detected matching portion on the display unit 14 or the like.

  Here, an effective feature amount will be described. In the case of a very short sentence, that is, a sentence with a small number of words, there is a high possibility that it is a different sentence despite having the same word length ratio fn.

  Therefore, it is effective that the number of words equal to or greater than the predetermined number of words, that is, the number (L) of the word length ratio fn included in the array be equal to or greater than the predetermined number as the characteristic amount for comparison. About L which makes a feature-value effective, 7 or more are preferable by an experimental result, More preferably, it is 10 or more.

  It should be noted that the ratio fn between two adjacent word lengths is effective as a feature amount for a document image described in a language having a different word length for each word. This is a document image written in a Latin language in which sentences are written using alphabets.

  In contrast, kanji, hiragana and katakana used in Japanese, Chinese, etc. are approximately the same size for each character, and it is difficult to clearly divide each word, so the pixel density is calculated for each character. Thus, it is effective to use this as a feature amount.

  FIG. 8 is a schematic diagram for explaining the pixel density. The pixel density is a ratio in which a circumscribed rectangle circumscribing one character is detected, and the number of pixels constituting one character included in the circumscribed rectangle occupies the total number of pixels constituting the entire circumscribed rectangle. As shown in FIG. 8, the calculation of the pixel density will be described using a single “I” character as an example.

  In the example of FIG. 8, first, the circumscribed rectangle 30 circumscribing the Chinese character “I” is extracted, and the width w (number of pixels) and height h (number of pixels) of the circumscribed rectangle 30 are counted. Next, the number of pixels B of the black pixels 31 constituting one character Kanji “I” included in the circumscribed rectangle 30 is counted. Since the pixel density is the ratio of the number B of black pixels constituting one character to the total number of pixels constituting the entire circumscribed rectangle as described above, when the pixel density is f, f = B / (w × h).

  Such a pixel density f is calculated for each character, and an array in which a plurality of pixel densities f are arranged in the order of characters on the text line is extracted as a feature amount.

  When an array of pixel density f is extracted for each line, an array of pixel density f is extracted for all lines included in one image data, and a feature amount of one image data is configured by all the arrays.

  When searching for document image data written in Japanese or Chinese, except that the feature amount is an array of pixel density f, a search process using the array of word length ratios fn as the feature amount Similarly, a search can be performed.

According to the present invention, the following effects can be obtained.
Compared to the case where the number of characters between punctuation marks is searched for as a feature amount as in the prior art, the same feature amount is obtained even in different sentences by using the ratio of the word lengths of two adjacent words as a feature amount. Since the possibility is low, the search accuracy of the document image data can be improved.

  Further, since the word length is not erroneously detected even if the document image data is read at a relatively low resolution, the low resolution document image data can be used, the processing speed related to the search processing is increased, and the document image data The storage capacity for storing can also be reduced.

  Further, similar image data can be searched from registered image data even if the input document image data and the registered document image data do not completely match the feature amount.

  It should be noted that each block of the image search apparatus 10, in particular, the input unit 12, the image search unit 13, the display unit 14, and the registered image storage unit 15 may be configured by hardware logic, or the CPU may be It may be realized by software (image search program).

That is, the image search apparatus 10 includes a central processing unit (CPU) that executes instructions of a control program that realizes each function, and a ROM (read only) that stores the program
memory), a RAM (random access memory) for expanding the program, and a storage device (recording medium) such as a memory for storing the program and various data. An object of the present invention is a recording medium in which program codes (execution format program, intermediate code program, source program) of a control program of the image search apparatus 10 which is software for realizing the above-described functions are recorded so as to be readable by a computer. Can also be achieved by supplying the program to the image search apparatus 10 and reading and executing the program code recorded on the recording medium by the computer (or CPU or MPU).

  Examples of the recording medium include tapes such as magnetic tapes and cassette tapes, magnetic disks such as floppy (registered trademark) disks / hard disks, and disks including optical disks such as CD-ROM / MO / MD / DVD / CD-R. Card system such as IC card, IC card (including memory card) / optical card, or semiconductor memory system such as mask ROM / EPROM / EEPROM / flash ROM.

  Further, the image search apparatus 10 may be configured to be connectable to a communication network, and the program code may be supplied via the communication network. The communication network is not particularly limited. For example, the Internet, intranet, extranet, LAN, ISDN, VAN, CATV communication network, virtual private network, telephone line network, mobile communication network, satellite communication. A net or the like is available. Also, the transmission medium constituting the communication network is not particularly limited. For example, even in the case of wired such as IEEE 1394, USB, power line carrier, cable TV line, telephone line, and ADSL line, infrared rays such as IrDA and remote control, Bluetooth ( (Registered trademark), 802.11 wireless, HDR, mobile phone network, satellite line, terrestrial digital network, and the like can also be used. 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 present invention can be implemented in various other forms without departing from the spirit or main features thereof. Therefore, the above-described embodiment is merely an example in all respects, and the scope of the present invention is shown in the claims, and is not restricted by the text of the specification. Further, all modifications and changes belonging to the scope of the claims are within the scope of the present invention.

DESCRIPTION OF SYMBOLS 10 Image search device 12 Input part 13 Image search part 14 Display part 15 Registered image memory | storage part 130 Pre-processing part 131 Feature extraction part 132 Search part

Claims (6)

An image search apparatus for searching document image data similar to input document image data from previously registered document image data,
A word contained in the input document image data is detected and divided into words, and for each two adjacent words, a ratio of the word lengths of the two words is calculated, and the calculated ratio of the word lengths is arranged as a word sequence. A feature amount extraction unit that extracts an array of word length ratios arranged in order as feature amounts of input document image data;
A registered image storage unit that stores the registered document image data and the feature amount of the registered document image data in association with each other;
Based on the feature amount of the registered document image data and the feature amount of the input document image data generated by the feature amount extraction unit, the input is performed from the registered document image data. A search unit for searching for image data similar to the document image data,
An image search apparatus comprising: a display unit configured to display document image data similar to the input document image data among the registered document image data based on a search result by a search unit.   The image search apparatus according to claim 1, wherein the feature amount extraction unit calculates a ratio of the word lengths using the number of pixels indicating the length of a region constituting the word as the word length. The registered image storage unit stores the word length ratio array in association with the positional information of the word length ratio array in the registered document image data,
The search unit detects a portion in the searched document image data that matches the word length ratio array of the input document image data,
The image search apparatus according to claim 1, wherein the display unit displays the detected matched portion so as to be distinguishable from other portions. An image search apparatus for searching document image data similar to input document image data from previously registered document image data,
Characters included in the input document image data are detected, character division is performed, a circumscribed rectangle circumscribing one character is detected, and a pixel density which is a ratio of pixels constituting the character in the circumscribed rectangle is calculated. A feature amount extraction unit that extracts an array of pixel densities obtained by arranging the calculated pixel densities along the word arrangement order as feature amounts of input document image data;
A registered image storage unit that stores the registered document image data and the feature amount of the registered document image data in association with each other;
Based on the feature amount of the registered document image data and the feature amount of the input document image data generated by the feature amount extraction unit, the input is performed from the registered document image data. A search unit for searching for image data similar to the document image data,
An image search apparatus comprising: a display unit configured to display document image data similar to the input document image data among the registered document image data based on a search result by a search unit.   An image search program for causing a computer to function as the image search device according to claim 1.   A computer-readable recording medium on which an image search program for causing a computer to function as the image search device according to claim 1 is recorded.

Images