JP2007102674A - Character image extraction device, image file generation device, character image extraction method and computer program - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To successfully extract an image only with a character part from a scan image even when the scan image is blurred. <P>SOLUTION: An image forming apparatus 1 is provided with a binary image generation part 172 which converts a character area image GM of a character area which is an area in which character images are aligned in an original image GA0 into a binary image GN on the basis of a first threshold, a character recognition processing part 173 which performs character recognition processing to the binary image GN, a character recognition ratio calculation part 174 which calculates character recognition ratio Rc of characters represented by the binary image GN, a character background image output part 179 which outputs the binary image GN as an extraction result of the character image of the original image GA0 when the character recognition ratio Rc of the binary image GN exceeds a second threshold and a reexecution instruction part 176 which controls the binary image generation part 172 so as to change the first threshold to reexecute conversion of the binary image GN when the character recognition ratio Rc of the binary image GN is less than the second threshold. <P>COPYRIGHT: (C)2007,JPO&INPIT

Description

  The present invention relates to an apparatus and method for extracting an image of a character part from an image of a document read by a scanner or the like.

  Scanners that read an image drawn on a paper document and convert it into electronic data are widely used. The scanner is not only used as a peripheral device such as a personal computer and a workstation, but also incorporated in a device such as a fax terminal, a copying machine, and an MFP (Multi Function Peripherals).

  An original may include an image representing characters (hereinafter referred to as “character image”). When generating electronic data of a document including a character image, it is necessary to reproduce the characters of the character image clearly. Therefore, a method as described in Patent Document 1 has been proposed.

  According to the method described in Patent Document 1, characters in transmission image data are recognized by character recognition means, the character recognition rate is compared with a preset setting value, and the character recognition rate is set to a set value. If it falls below, the transmission resolution is raised from the default value. As a result, when thin characters or small characters that are difficult to recognize as characters are written on the document for transmission, the resolution of the image data for transmission is automatically changed, and a suitable image can be transmitted. .

  In addition, in order to reduce the size of the image file of the read original, the entire image of the original is divided into a character image area and a photo area, and the image data is compressed by a compression method suitable for each area. The method is used. One typical file format that employs such a method is PDF (Portable Document Format) developed by Adobe Systems (see Non-Patent Document 1).

In addition, there are proposed methods for automatically determining the position of a character image to be subjected to character recognition processing from images of a document as described in Patent Documents 2 and 3.
JP 2001-103311 A JP-A-5-252288 JP-A-7-168911 "What is Adobe PDF?-PDF beginner's edition", Adobe Systems, May 14, 2005, Internet <URL: http://www.adobe.co.jp/products/acrobat/adobepdf13.html >

  By the way, in documents such as books, magazines, and brochures, a character image in which a color or a pattern is decorated on the background of the character is often seen. When a document including such a character image is scanned as a document, a clear character image may not be obtained because the contrast between the character and its background is small. In addition, if the background of the document has a color or pattern, the background removal process may be performed excessively on the character image, resulting in thin character lines and a clear character image may not be obtained. .

  When converting an image of a document into electronic data, it is required that characters appear clearly in the character image. In addition, when converting image data of a document image into a file of a format such as PDF, it is necessary to extract an image of only the character portion excluding the background portion from the character image. If the image is unclear, it cannot be extracted successfully. However, the conventional method described in Patent Document 1 cannot generate a black and white character image from an unclear character image. In addition, changing the scanning conditions and rescanning the document is time consuming and troublesome for the user, and the rescanning is not always successful.

  In view of such problems, an object of the present invention is to enable an image of only a character portion to be successfully extracted from a character image even if the character image obtained by a scanner or the like is unclear.

  A character image extraction device according to the present invention is a character image extraction device that extracts a character image from an input image input by an image input device, wherein one or a plurality of character images of the input images are extracted. A binary image conversion unit that executes a conversion process for converting an image of a character area that is a lined area into a binary image based on a first threshold value, and the converted binary image of the character area Character recognition processing means for executing character recognition processing and calculating a character recognition rate, which is a ratio of characters recognized by the character recognition processing among characters represented by the binary image of the character region Means, and when the character recognition rate of the binary image of the character area exceeds a second threshold, the output means for outputting the binary image as a character image extraction result of the input image, and the character area Before binary image A reconversion control unit that controls the binary image conversion unit to change the first threshold and re-execute the conversion process for the character region when the character recognition rate is lower than the second threshold. It is characterized by having.

  Preferably, the re-conversion control unit increases the first threshold value and re-executes the conversion process when the background density of the image of the character region in the input image exceeds a third threshold value. If the background density is lower than the third threshold value, the binary image conversion means is controlled so that the first threshold value is lowered and the conversion process is executed again. To do.

  “Character recognition” generally means that a character in an image is recognized and converted into a character code such as an ASCII code. However, in the present invention, it is not necessary to be able to specify one character represented by an image, You can narrow down to several candidates. For example, as a result of character recognition of an image representing the character “B”, if it is found that it is any of “B”, “2”, or “Z”, the character recognition of the image has been completed. To do.

  In the present invention, the “character” includes not only kanji, hiragana, katakana, and alphabet, but also numbers and symbols.

  According to the present invention, even if a character image obtained by a scanner or the like is unclear, an image of only the character portion can be successfully extracted from the character image.

  1 is a diagram showing an example of a network configuration having an image forming apparatus 1 according to the present invention, FIG. 2 is a diagram showing an example of a hardware configuration of the image forming apparatus 1, and FIG. 3 is a functional configuration of the image forming apparatus 1. It is a figure which shows an example.

  An image forming apparatus 1 shown in FIG. 1 is an image processing apparatus that integrates various functions such as copying, scanning, faxing, network printing, document server, and file transfer. There is also a device called a multifunction device or MFP (Multi Function Peripherals).

  The image forming apparatus 1 is installed in offices or offices such as businesses, public facilities such as schools or libraries, stores such as convenience stores, and other various places and can be shared by a plurality of users. Further, it can be connected to a terminal device 2 such as a personal computer or a workstation via a communication line 3. As the communication line 3, the Internet, a LAN, a public line, a dedicated line, or the like is used.

  As shown in FIG. 2, the image forming apparatus 1 includes a CPU 10a, a RAM 10b, a ROM 10c, a hard disk 10d, a control circuit 10e, an operation panel 10f, a scanner 10g, a printing device (engine) 10h, a modem 10j, a network interface 10k, and the like. Composed.

  The scanner 10g is an apparatus that optically reads an image such as a photograph, a character, a picture, or a chart drawn on a document sheet (hereinafter simply referred to as “document”) and converts it into electronic data. The read image data is stored in the RAM 10b, and is subjected to various processes and filed as described later.

  The printing apparatus 10h is an apparatus that prints an image read by the scanner 10g or an image transmitted from the terminal device 2 or the like on a sheet. In the case of a color MFP, an image is printed using toners of four colors of yellow, magenta, cyan, and black.

  The operation panel 10f includes an operation unit and a display unit. A numeric keypad or the like is used as the operation unit, and a liquid crystal display or the like is used as the display unit. By operating the operation unit, the user gives an instruction to start or stop the processing to the image forming apparatus 1 and designates processing conditions such as a data destination, a scanning condition, or an image file format, Various other items can be specified. The display unit includes a screen for giving a message or an instruction to the user, a screen for inputting the type of processing and processing conditions desired by the user, and a screen showing the result of the processing executed by the image forming apparatus 1 Etc. are displayed. When a touch panel is used as the operation panel 10f, the touch panel serves as both an operation unit and a display unit. As described above, the operation panel 10 f serves as a user interface for a user who operates the image forming apparatus 1. Note that an application program and a driver for giving a command to the image forming apparatus 1 are installed in the terminal device 2. Therefore, the user can remotely operate the image forming apparatus 1 with the terminal device 2.

  The modem 10j has a built-in NCU (Network Control Unit), and is connected to another fax terminal via an analog public line to perform data control based on a facsimile protocol, fax data modulation / demodulation, and the like. It is also possible to connect to the Internet via an access point of an Internet provider (ISP) and send / receive a file to / from the terminal device 2 or the like by e-mail or FTP (File Transfer Protocol). The network interface 10k is a NIC (Network Interface Card), and connects the image forming apparatus 1 to the terminal device 2 or the like via a LAN, and transmits and receives files.

  The control circuit 10e is a circuit for controlling devices such as the hard disk 10d, the operation panel 10f, the scanner 10g, the printing device 10h, the modem 10j, and the network interface 10k.

  The hard disk 10d includes a preprocessing unit 101, a brightness calculation unit 102, a smoothing processing unit 103, a region detection unit 104, a region image extraction unit 105, a background density detection unit 106, a character background image generation unit 107, as shown in FIG. A program, data, and the like for realizing functions such as the image compression processing unit 108 and the file generation unit 109 are stored. These programs are read into the RAM 10b as necessary, and the programs are executed by the CPU 10a. Some or all of these programs or data may be stored in the ROM 10c. Alternatively, part or all of the functions shown in FIG. 3 may be realized by the control circuit 10e.

  In addition, the CPU 10a detects the contents of the operation by the user (that is, the key or button pressed by the user), controls the hardware constituting the image forming apparatus 1, and generates e-mail data. 1 basic processing is performed. These processes are performed based on an operating system program stored in the hard disk 10d.

  FIG. 4 is a diagram illustrating an example of a document image GA0, and FIG. 5 is a diagram illustrating an example of a character image having a defect.

  Next, functions and processing contents of each unit of the image forming apparatus 1 shown in FIG. 3 will be described. The preprocessing unit 101 converts a document image (hereinafter referred to as “document image GA0”) input by the scanner 10g into a resolution (resolution conversion) corresponding to the processing after the lightness calculation unit 102. In addition, a process for removing the color of the background portion of the document image GA0 (background removal process) is performed. The “background portion of the document image GA0” means an area where there is no content (object). For example, in the case of a manuscript image GA0 as shown in FIG. 4, a rectangular area with a character string image “New Digital MFP XYZ-2005 Series”, “The price difference of digital machines can be expected.” A portion other than the three rectangular regions of the rectangular region where the character string image is located and the rectangular region where the photograph of the MFP is arranged at the center is the background portion of the document image GA0.

  The document image GA0 includes images of various types of objects such as character images, photo images, illustration images, or graph images. Hereinafter, an area including a character image is referred to as a “character area RM”, and an area including an image of an object other than a character is referred to as a “non-character area RH”. For example, the document image GA0 of FIG. 4 has two character areas RM and one non-character area RH. Hereinafter, the document image GA0 subjected to the resolution conversion process and the background removal process is referred to as “document image GA0 ′”.

  The brightness calculation unit 102 generates the brightness image GA1 by calculating the brightness of each pixel of the document image GA0 '. The smoothing processing unit 103 performs a smoothing process on the lightness image GA1 to remove noise from the lightness image GA1, correct a contour, and the like. Hereinafter, the lightness image GA1 that has been subjected to the smoothing process is referred to as a “lightness image GA1 '”.

  The area detection unit 104 determines the positions of the character area RM and the non-character area RH included in the document image GA0 'by performing a labeling process on the lightness image GA1'. In general, when a labeling process is performed on a character image, many irregular thin lines appear, and when a labeling process is performed on an image such as a photograph, a large white or black lump appears. Based on such characteristics, the positions of the character region RM and the non-character region RH are determined, and region position data 81 indicating the determination result is generated. In the case of FIG. 4, region position data 81 is generated for each of the character regions RMa, RMb and the non-character region RHa.

  The area image extraction unit 105 extracts the image of the character area RM and the image of the non-character area RH from the document image GA0 ′ based on the detection result by the area detection unit 104. When a document image GA0 as shown in FIG. 4 is input by the scanner 10g, each image of the character areas RMa and RMb and the non-character area RHa is extracted. Hereinafter, the extracted image of the character region RM is referred to as “character region image GM”, and the extracted image of the non-character region RH is referred to as “non-character region image GH”. The character area image GM includes one or more character images. For example, the character region RMb includes approximately 120 character images including characters such as numbers and punctuation marks.

  The character background image generation unit 107 separates the character region image GM into an image with only characters (that is, an image with the background deleted) and an image with only the background in order to increase the compression rate of the character region image GM. That is, the character background image generation unit 107 generates an image of only characters and an image of only the background from the character region image GM.

  By the way, if the character image included in the character region image GM has a low contrast between the character and its background (character background) or if excessive background removal processing is performed, an image of only the character is generated. The characters may be crushed without being separated from the background as shown in FIG. 5A, or the characters may be faded as shown in FIG. Therefore, the character background image generation unit 107 executes correction (optimization) processing as necessary so that the characters become clearer when generating an image of only characters. Hereinafter, the character background image generation unit 107 will be described in detail.

  6 is a diagram illustrating an example of the configuration of the character background image generation unit 107, FIG. 7 is a diagram illustrating an example of grouping of character images included in the character region RMb, and FIG. 8 is a diagram illustrating an example of the density threshold table TL1. FIG. 9 is a diagram showing an example of a defective binary image GN and a good binary image GN.

  As shown in FIG. 6, the character background image generation unit 107 includes a character region image division unit 170, a character background density detection unit 171, a binary image generation unit 172, a character recognition processing unit 173, a character recognition rate calculation unit 174, two A value image pass / fail judgment unit 175, a re-execution command unit 176, a character image integration unit 177, a background image generation unit 178, a character background image output unit 179, a density threshold value table TL1, and the like.

  The character area image dividing unit 170 groups a plurality of character images arranged in the horizontal direction or the vertical direction when the character area image GM includes character images of a large number of characters. For example, about 1 to 10 character images are grouped. In the case of English sentences, character images may be grouped for each word (that is, between spaces). For horizontal writing, character images arranged in the horizontal direction are grouped, and for vertical writing, character images arranged in the vertical direction are grouped. Whether it is horizontal writing or vertical writing can be determined based on the characteristics of the arrangement of the character images. By grouping in this way, a plurality of blocks made up of one or more character images are generated. Hereinafter, the generated image of the block is referred to as “block image GR”. For example, the character area image GM of the character area RMb (see FIG. 4) is divided into a plurality of block images GR (GR1, GR2,...) As shown in FIG. For each block image GR, character string position data 82 indicating the position in the character region RM is generated.

  The character background density detector 171 detects the density of the background of the block image GR, that is, the density of the background portion of the character, for example, as follows. Usually, since the character portion is located at the center of the block image GR, the edges (upper and lower and left and right ends) of the block image GR are often the background (background) of the character. Therefore, the density of the background portion of the block image GR is detected by examining the pixels near the edge of the block image GR. Hereinafter, the detected density is referred to as “character base density Hd”.

  The binary image generation unit 172 generates a binary image GN of each block image GR obtained by the character area image division unit 170. In the present embodiment, the binary image GN of the block image GR is generated as follows.

  The density of each pixel of the block image GR is calculated. In the present embodiment, the density is represented by 8 bits. If the calculated density is equal to or higher than the density threshold β, the value of the pixel is “1”, and if the calculated density is less than the density threshold β, the value of the pixel is “0”. Thus, by converting the value of each pixel into either “0” or “1”, a binary image GN is generated. The density threshold β is defined by the density threshold table TL1 shown in FIG. In the example of FIG. 8, the initial setting value is “128”.

  The character recognition processing unit 173 performs character recognition processing on the binary image GN of each block image GR, and determines the character represented by the character image included in the binary image GN. At this time, the number of character images included is also counted.

  The character recognition rate calculation unit 174 counts the number of character images in which characters can be identified (recognized) by the character recognition processing unit 173 among the character images included in the binary image GN. Then, a ratio of the number of recognized character images to the total number of character images, that is, a character recognition rate Rc is calculated. For example, when 9 characters are recognized as a result of executing the character recognition process on the binary image GN representing 10 characters “the price difference of the digital machine”, Rc = 90%.

  When the character recognition rate Rc of the binary image GN is equal to or higher than the recognition threshold value α, the binary image quality determination unit 175 determines that the binary image GN clearly shows characters and is good. When it is less than the recognition threshold value α, it is determined that the binary image GN is defective because it does not clearly represent characters. According to the binary image pass / fail determination unit 175, the binary image GN is a character image that is thinly faded as shown in FIG. When the ratio is large, it is determined that it is defective. As shown in FIG. 9C, when the ratio of character images in which characters appear clearly is large, it is determined that the characters are good. 9A to 9C, “◯” means that the character represented by the character image has been recognized by the character recognition processing unit 173, and “X” means that the character has not been recognized. .

  The re-execution command unit 176 instructs the binary image generation unit 172 to regenerate the binary image GN determined to be defective. Then, the binary image generation unit 172 discards the data of the binary image GN and generates the binary image GN again.

  However, if the character background density Hd of the original block image GR of the binary image GN is greater than or equal to the dark boundary threshold γ, the background portion in the binary image GN is part of the character last time. It is considered that the character is crushed due to being erroneously detected, so the density threshold value β is raised and regenerated. In the present embodiment, as shown in FIG. 8, five levels of density threshold value β are defined in advance in the density threshold value table TL1. When the binary image GN is generated again, the density threshold value β is increased by adopting a value indicated by a level one level higher than the level when the binary image GN of the block image GR was previously generated. . For example, if the binary image GN was previously generated with the level 3 density threshold β, the binary image GN is regenerated based on the level 4 density threshold β.

  On the other hand, when the character background density Hd is less than the dark boundary threshold γ, it is considered that the character portion in the binary image GN was faint in the previous time, so the density threshold value one level lower than the previous time. The value is lowered to β to regenerate the binary image GN.

  The regenerated binary image GN is subjected to character recognition processing again by the character recognition processing unit 173, the character recognition rate calculation unit 174, and the binary image pass / fail determination unit 175, and the character recognition rate Rc is calculated. And pass / fail is discriminated. As a result, when it is determined that it is defective, the binary image GN is regenerated by raising and lowering the level by another level. Hereinafter, this is repeated until it is determined to be good. However, if the density threshold value β cannot be increased or decreased any more, the last generated binary image GN is regarded as good, and the processing after the character image integration unit 177 described below is performed. Use for.

  The character image integration unit 177 arranges the good binary images GN of the respective block images GR, which are the divided images of the character region image GM, generated by the binary image generation unit 172, according to the respective character string position data 82, Merge into one image. This integrated image is obtained by removing the background portion from the character region image GM and extracting only the character portion. Therefore, hereinafter, an image generated by integrating these binary images GN will be referred to as a “character extraction image GT”.

  The background image generation unit 178 generates a background image GB that is an image of only the background portion of the character region RM. That is, an image having the same shape as the character region image GM and the same color or pattern as the background portion of the character region image GM is generated. The color or pattern can be detected by examining pixels near the edge of the character area image GM. Since the background image GB is used as a character background, the tone or color of the background image GB is set to be higher than that of the detected color or pattern in order to make the characters easier to see when reproducing the document image GA0. You may correct | amend so that it may become low.

  Alternatively, the background image GB may be generated by generating a background image for each block image GR that is a divided image of the character area image GM and arranging them in the order of arrangement of the original block images GR.

  The character background image output unit 179 outputs the generated data of the character extraction image GT and the background image GB to the image compression processing unit 108 (see FIG. 3).

  FIG. 10 is a flowchart for explaining an example of the flow of the character background image generation process, FIG. 11 is a flowchart for explaining an example of the flow of the binary image generation process, and FIG. 12 is a diagram showing an example of the configuration of the file FL.

  Here, the entire character background image generation unit 107 is taken as an example in which the character extraction image GT and the background image GB are generated by separating the character region image GM of the character region RMb of FIG. 4 into a character portion and a background portion. A typical process flow will be described with reference to the flowcharts of FIGS.

  The character background image generation unit 107 has about 1 to 10 character images of each character “I can expect the price difference of digital machines” included in the character region image GM of the character region RMb as shown in FIG. Are divided into groups of character images and divided into block images GR (GR1, GR2,...) (# 11 in FIG. 10).

  Steps # 12 to # 14 are performed on the first block image GR1. That is, the character string position data 82 is obtained by determining where the block image GR1 is in the character area RMb (# 12), and the background density (character background density Hd) of the block image GR1 is detected (# 12). # 13).

  A binary image GN of the block image GR1 is generated by a procedure as shown in FIG. 11 (# 14). That is, first, based on the density threshold table TL1 (see FIG. 8), the density threshold β is reset and set to an initial level value (# 21). Each pixel of the block image GR1 is binarized using the density threshold β as a threshold to generate a binary image GN (# 22). Then, by executing the character recognition process, the character recognition of each character image in the binary image GN is performed, and the character recognition rate Rc is calculated (# 23).

  When the calculated character recognition rate Rc is equal to or greater than the recognition threshold value α (No in # 24), it is considered that the character appears clearly in the binary image GN, and the binary image GN generated this time is blocked. Adopted as a good binary image GN of the image GR1 (# 34).

  On the other hand, when the character recognition rate Rc is less than the recognition threshold value α (Yes in # 24), the density threshold value β is changed as follows to generate a binary image GN so that a clearer character image appears. Try again.

  When the character background density Hd of the block image GR1 is greater than or equal to the thin boundary threshold γ (Yes in # 25), a binary image GN in which characters are not clearly displayed due to the dark background portion of the block image GR1 is obtained. It is thought that it was generated. Therefore, based on the density threshold table TL1, the level is increased by 1 to increase the density threshold β (# 26), and the binary image GN is generated again with the density threshold β (# 27). Then, character recognition processing is executed for the regenerated binary image GN (# 28), and if the character recognition rate Rc is equal to or greater than the recognition threshold α (No in # 29), the latest (that is, the current regenerated image) The binary image GN is adopted as a good binary image GN of the block image GR1 (# 34).

  When the character background density Hd of the block image GR1 is less than the thin boundary threshold γ (No in # 25), it is considered that the characters appearing in the binary image GN are faint. Therefore, the level is decreased by one to reduce the density threshold β (# 30), and the binary image GN is generated again with the density threshold β (# 31). Then, the character recognition process is executed for the regenerated binary image GN (# 32), and if the character recognition rate Rc is equal to or greater than the recognition threshold α (No in # 33), the regenerated binary image GN this time. Is adopted as a good binary image GN of the block image GR1 (# 34).

  If the character recognition rate Rc is still less than the recognition threshold α (Yes in # 29 or Yes in # 33), the level of the density threshold β is further increased and decreased to generate the binary image GN and perform the image recognition process again. However, if the level cannot be further increased (No in # 29) or cannot be decreased (No in # 33), the binary image GN generated last is used as a good binary image of the block image GR1. It is regarded as GN (# 34).

  Returning to FIG. 10, when the binary image GN of the block image GR1 is obtained, the binary images GN of the remaining block images GR (GR2, GR3,...) Are sequentially generated. The generation procedure is the same as that for generating the binary image GN of the block image GR1 described above (# 12 to # 14).

  When good binary images GN of all the block images GR are generated (Yes in # 15), these binary images GN are arranged in the original arrangement order based on the character string position data 82 obtained in Step # 12. A character extraction image GT is generated by integrating the images into one image (# 16).

  The color or pattern of the background portion of the block image GR is detected, and the background image GB is generated (# 17).

  Returning to FIG. 3, the image compression processing unit 108 converts the image of only the character portion of the character region RM, that is, the image data of the binary image GN, and the image of only the background portion, that is, the image data of the background image GB, to the character background image generation unit 107. If acquired from the above, these image data are compressed. Since the character extraction image GT is a monochrome bitmap image, it is desirable to compress it using the G4 compression method (MMR compression method) or the like. Since the background image GB is a color or pattern image, it is desirable to compress the background image GB by a compression method such as GIF or JPEG. Hereinafter, the image data of the compressed character extraction image GT and the image data of the background image GB are referred to as “compressed character image data DTA” and “compressed background image data DTB”, respectively.

  The image compression processing unit 108 also compresses image data in the non-character region RH. However, since a large-sized image such as a photograph is arranged in the non-character region RH, the image is compressed so as to be reduced in size by an irreversible compression method such as JPEG. Alternatively, the image type may be determined for each non-character area image GH, and a compression method such as JPEG or GIF may be used properly. Hereinafter, the image data of the non-character area RH compressed by the image compression processing unit 108 is referred to as “compressed non-character image data DTC”.

  The file generation unit 109 uses the compressed data obtained by the image compression processing unit 108 and the position information of each image, for example, as a file FL for reproducing the document image GA0 of the document read by the scanner 10g. Generate as follows.

  Compressed character image data DTA, compressed background image data DTB, compressed non-character image data DTC, and region position data 81 for each area constituting the document image GA0 are prepared. Further, file attribute data 83 indicating the attribute of the file FL to be generated is generated and prepared. A part of the attribute (for example, a file name) may be specified by the user.

  The compressed character image data DTA, the compressed background image data DTB, and the region position data 81 in the same character region RM are associated with each other, and the compressed background image data DTB in the same non-character region RH and the region position data 81 are associated with each other. Then, the compressed character image data DTA, the compressed background image data DTB, the compressed non-character image data DTC, the region position data 81, and the attribute data 83 are integrated into one file. In this way, the file FL is generated.

  For example, an original image GA0 as shown in FIG. 4 is input by the scanner 10g, and the original image GA0 is processed by the preprocessing unit 101 or the image compression processing unit 108 of FIG. Compressed character image data DTAa, compressed background image data DTBa, and region position data 81a are obtained. As character region RMb data, compressed character image data DTAb, compressed background image data DTBb, and region position data 81b are obtained, and non-characters are obtained. It is assumed that compressed non-character image data DTCc and region position data 81c are obtained as data of region RHa. In such a case, the file generation unit 109 generates the attribute data 83, arranges and associates the data for each area, and generates a file FL having a file configuration as shown in FIG.

  If there are a plurality of areas of the same type of object, these areas may be integrated into one and processing for reducing the total number of objects may be performed.

  The generated file FL is stored in the hard disk 10d. Alternatively, it is transferred to the terminal device 2 or the like via the communication line 3 by e-mail or FTP. The file FL may be generated in accordance with an existing format such as Acrobat PDF.

  The generated image of the file FL is reproduced in the terminal device 2 or the like as follows. The terminal device 2 activates application software corresponding to the file FL (for example, Adobe Acrobat's acrobat reader if the file FL is a PDF file), and loads the file FL into the RAM. A character image is reproduced based on the compressed character image data DTA in the file FL, and the image is arranged at a position indicated by the region position data 81 associated therewith. When the compressed background image data DTB is associated, a background image is reproduced based on the compressed background image data DTB, and a character image is superimposed on the background image in a transparent format. Similarly, with respect to the compressed non-character image data DTC, an image such as a photograph may be reproduced based on the data, and the image may be arranged based on the position indicated by the region position data 81 associated with the data. In this way, the document image GA0 is reproduced.

  FIG. 13 is a flowchart illustrating an example of the overall processing flow of the image forming apparatus 1 when executing a scan job. Next, the overall processing flow of the image forming apparatus 1 when the original is scanned to generate electronic data of the original image will be described with reference to the flowchart of FIG.

  The user sets a document to be converted into electronic data on the document table of the image forming apparatus 1, and operates the operation panel 10 f to give a scan command to the image forming apparatus 1.

  Then, the image forming apparatus 1 scans the set original and inputs the original image GA0. That is, the data of the document image GA0 is acquired (# 1 in FIG. 13). Then, processing for generating the file FL of the document image GA0 is started as follows. That is, first, resolution conversion processing and background removal processing are performed on the document image GA0 (# 2). Thereby, the document image GA0 'is obtained. A brightness image GA1 is generated by calculating the brightness of each pixel of the document image GA0 '(# 3), and a smoothing process is performed on it (# 4). Thereby, the brightness image GA1 'is obtained.

  By performing a labeling process on the brightness image GA1 ', the character area RM and the non-character area RH in the document image GA0' are detected (# 5). At this time, area position data 81 indicating the positions of these areas is generated.

  From the character region image GM in the character region RM, an image of only characters, that is, a binary image GN, and an image of only the background, that is, a non-character region image GH are generated (# 6). At this time, for the binary image GN, optimization processing is performed as necessary so that the characters become clear. The procedure of such processing is as described above with reference to FIGS. Note that the non-character area image GH of the non-character area RH may be subjected to processing for correcting a photograph or an illustration using a known method.

  The image data of the character area image GM, background image GB, and non-character area image GH are compressed to generate compressed character image data DTA, compressed background image data DTB, and compressed non-character image data DTC (# 7).

  The file FL is generated by integrating the compressed character image data DTA, the compressed non-character image data DTC, the compressed background image data DTB, the region position data 81 of each region, the attribute data 83, and the like (# 8). ).

  According to the present embodiment, even if the character image included in the original image GA0 obtained by scanning contains background decoration or the character image is subjected to excessive background removal processing, the character image becomes unclear. , It is possible to successfully extract an image of only the character portion from the character image.

  In addition, since the optimization process is performed after the determination of the character region and the character background density Hd, it is possible to suppress the processing time and the use of resources such as memory.

  In this embodiment, all the functions of the respective parts in FIGS. 2 and 3 are integrated, such as a function of reading an image of a document, a function of converting an image format, a function of correcting an image, and a function of filing image data. Although the image forming apparatus 1 has been described as an example, the functions of the respective units in FIG. 3 can be realized by being distributed to a plurality of apparatuses. For example, the function of reading an image of a document may be realized by a scanner for a personal computer, and the others may be realized by a personal computer. In this case, a scanner driver and a computer program for executing the processing described with reference to FIGS. 10, 11, and 13 are installed in the personal computer. Alternatively, the image data of the original image GA0 read by the image forming apparatus 1 may be transferred to the terminal apparatus 2 so that the terminal apparatus 2 executes the processing described with reference to FIGS.

  In the present embodiment, the density threshold β is adjusted at five levels, but may be adjusted at a finer level.

  In this embodiment, the case where an image of a document read by a scanner is converted into a file has been described as an example. However, the present invention can also be applied to a case where an image acquired by another type of image reading apparatus is converted into a file. is there. For example, the present invention can be applied to file an image of a document shot with a digital camera.

  The text data of the character string represented by the character region image GM obtained when the character recognition process is performed on the character region image GM may be included in the file FL.

  In the present embodiment, the character area RM and the non-character area RH are automatically detected. However, the user may be able to specify them by dragging with the mouse.

  In addition, the configuration of the entire image forming apparatus 1 or each unit, processing contents, processing order, table contents, and the like can be appropriately changed in accordance with the spirit of the present invention.

  The present invention is particularly suitable for generating a PDF file of a scanned document image in an image processing apparatus such as an MFP or a personal computer.

1 is a diagram illustrating an example of a network configuration having an image forming apparatus according to the present invention. 2 is a diagram illustrating an example of a hardware configuration of an image forming apparatus. FIG. 2 is a diagram illustrating an example of a functional configuration of an image forming apparatus. FIG. It is a figure which shows the example of a manuscript image. It is a figure which shows the example of the character image which has a malfunction. It is a figure which shows the example of a structure of a character background image generation part. It is a figure which shows the example of grouping of the character image contained in a character area. It is a figure which shows the example of a density | concentration threshold value table. It is a figure which shows the example of a binary image with a malfunction, and a favorable binary image. It is a flowchart explaining the example of the flow of a character background image generation process. It is a flowchart explaining the example of the flow of a binary image generation process. It is a figure which shows the example of a structure of a file. 10 is a flowchart illustrating an example of the overall processing flow of the image forming apparatus when executing a scan job.

Explanation of symbols

1 Image forming device (character image extraction device)
10g scanner (image input device, image file generation device)
104 area detection unit (character area detection means)
108 Image compression processing unit (data compression means)
109 File generation unit (image file generation means)
170 Character area image dividing unit (dividing means)
172 Binary image generation unit (binary image conversion means)
173 Character recognition processing unit (character recognition processing means)
174 Character recognition rate calculation unit (character recognition rate calculation means)
176 Re-execution command section (re-conversion control means)
177 Character image integration unit (integrated image generation means)
179 Character background image output unit (output means)
FL file (image file)
Hd Character background density GA0 Original image (input image)
GN binary image (character image)
GR block image GT character extraction image (integrated image)
Rc Character recognition rate RH Non-character area RM Character area α Recognition threshold (second threshold)
β concentration threshold (first threshold)
γ Dense boundary threshold (third threshold)

Claims (8)

A character image extraction device that extracts a character image from an input image input by an image input device,
Binary image conversion that executes conversion processing for converting an image of a character area, which is an area in which one or more character images are arranged, from the input image into a binary image based on a first threshold. Means,
Character recognition processing means for executing character recognition processing on the converted binary image of the character region;
A character recognition rate calculating means for calculating a character recognition rate, which is a ratio of characters recognized by the character recognition process among characters represented by the binary image of the character region;
An output unit that outputs the binary image as a character image extraction result of the input image when the character recognition rate of the binary image of the character region exceeds a second threshold;
When the character recognition rate of the binary image of the character area is lower than the second threshold value, the binary image conversion is performed such that the first threshold value is changed and the conversion process is performed again for the character area. Reconversion control means for controlling the means;
A character image extracting apparatus comprising: The re-conversion control means increases the first threshold value and re-executes the conversion process when the background density of the image of the character region in the input image exceeds a third threshold value. Controlling the value image converting means, and if the background density is lower than the third threshold value, the binary image converting means is controlled so as to lower the first threshold value and execute the conversion process again.
The character image extraction apparatus according to claim 1. A character image extraction device that extracts a character image from an input image input by an image input device,
A dividing unit that divides a character region, which is a region where a plurality of character images in the input image are arranged, into a plurality of blocks such that one or more character images are grouped;
Binary image conversion means for executing conversion processing for converting an image of the block of the input image into a binary image based on a first threshold;
Character recognition processing means for executing character recognition processing on the binary image of the converted block;
A character recognition rate calculating means for calculating a character recognition rate, which is a ratio of characters recognized by the character recognition process among characters represented by the binary image of the block;
When the character recognition rate of the binary image of the block is lower than a second threshold, the binary image conversion means is controlled so as to change the first threshold and re-execute the conversion process for the block. Re-conversion control means;
An integrated image generating means for generating an integrated image by arranging binary images in which the character recognition rate of each block of the character area exceeds the second threshold in the original arrangement order and integrating them into one image; ,
Output means for outputting the generated integrated image of the character region as a result of extraction of the character image of the input image;
A character image extracting apparatus comprising: When the background density of the image of the block exceeds a third threshold value, the re-conversion control means increases the first threshold value and re-executes the conversion process for the block. If the background density is lower than the third threshold, the binary image conversion means is controlled so as to lower the first threshold and re-execute the conversion process for the block.
The character image extraction device according to claim 3. An image file generation device that generates an image file of an input image input by an image input device,
A character area detecting means for detecting, from the input image, a character area in which one or more character images are continuously arranged and a non-character area in which an image other than characters is arranged;
A binary image conversion means for executing a conversion process of converting the detected image of the character region of the input image into a binary image based on a first threshold;
Character recognition processing means for executing character recognition processing on the converted binary image of the character region;
A character recognition rate calculating means for calculating a character recognition rate, which is a ratio of characters recognized by the character recognition process among characters represented by the binary image of the character region;
In the case where the character recognition rate of the binary image of the character area is lower than a second threshold value, when the background density of the character area exceeds a third threshold value, the first threshold value is increased and the character area is The binary image converting means is controlled so as to execute the conversion process again. When the background density of the character area is lower than the third threshold value, the first threshold value is lowered and the conversion process is performed for the character area. Re-conversion control means for controlling the binary image conversion means to re-execute,
Image data of the binary image of the character area determined that the character recognition rate exceeds the second threshold, image data of the image of the non-character area, each of the character area and the non-character area, Image file generation means for generating an image file by integrating the position data indicating the position in the input image and the background color data for reproducing the background color of the character area into one;
An image file generation device characterized by comprising: Data compression means for compressing image data of an image according to the type of object represented by the image;
The image file generation means generates the image file by integrating the image data of the binary image in the character area and the image data of the image in the non-character area compressed by the data compression means.
6. The image file generation means according to claim 5. A character image extraction method for extracting a character image from an input image input by an image input device,
Executing a conversion process of converting an image of a character area, which is an area in which one or more character images are arranged in the input image, into a binary image based on a first threshold;
Performing character recognition processing on the converted binary image of the character region;
Calculating a character recognition rate, which is a ratio of characters recognized by the character recognition process among characters represented by the binary image of the character region;
When the character recognition rate of the binary image of the character region exceeds a second threshold value, the binary image is output as a character image extraction result of the input image,
When the character recognition rate of the binary image of the character region is lower than the second threshold, the binary image conversion is performed so that the first threshold is changed and the conversion process is performed again for the character region. Control means,
A character image extraction method characterized by that. A computer program used in a computer for extracting a character image from an input image input by an image input device,
A conversion process of converting an image of a character area, which is an area in which one or more character images are arranged in the input image, into a binary image based on a first threshold;
A character recognition process for the binary image of the converted character region;
A calculation process for calculating a character recognition rate which is a ratio of characters recognized by the character recognition process among the characters represented by the binary image of the character region;
When the character recognition rate of the binary image of the character region exceeds a second threshold, an output process for outputting the binary image as a character image extraction result of the input image;
When the character recognition rate of the binary image of the character area is lower than the second threshold value, the binary image conversion is performed such that the first threshold value is changed and the conversion process is performed again for the character area. Control processing for controlling the means;
A computer program for causing a computer to execute.