JP2007122500A - Character recognition device, character recognition method and character data - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a character recognition device and a character recognition method, in which a character can be accurately recognized at high speed, and character data for the character recognition. <P>SOLUTION: The character recognition device for recognizing a character from optically scanned image data of a document comprises a font determination means 21 determining the font of the character, a character size determination means 22 determining the size of the character, character data 5 for recognizing characters, which is stored in conformation to fonts of characters and character sizes, and a character identification means 24 determining, based on the font determined by the font determination means and the character size determined by the character size determination means, a character code of the character in reference to the character data. <P>COPYRIGHT: (C)2007,JPO&INPIT

Description

  The present invention relates to a character recognition device, a character recognition method, or character data for character recognition for recognizing characters from image data of an optically scanned document.

There has been proposed a character recognition technique for digitizing a manuscript on which characters created by a word processor or the like are printed so that the information processing apparatus such as a computer can handle the document. In the character recognition technology, an original is read by an image scanner or the like to recognize characters, and the characters are converted into character codes such as alphanumeric characters, hiragana or kanji and stored.

  In a conventional character recognition device, a character pattern identification dictionary used for character recognition by averaging feature values of a plurality of representative fonts (Gothic, Mincho, serif, sans serif, and monospace) is used. I was making it. However, there is a limit to discriminating power in a character pattern in which feature amounts are averaged.

  For this reason, a character recognition technique for preparing an identification dictionary for each font has been proposed (see, for example, Patent Document 1). In such a technique, a font incorporated in a personal computer or the like is detected, and a standard pattern is created and registered for each character for each font. In the case of character recognition, characters are recognized from image data read by a scanner or the like using a registered standard pattern.

In addition, a feature amount related to the shape of a character in image data read by a scanner device or the like is extracted, a similarity with a feature amount for each font type prepared in advance is calculated, and a font type is identified based on the similarity. A technique has been proposed (see, for example, Patent Document 2). In such a technique, the outline information of the character is prepared in advance according to the type of the font. Therefore, if the font is identified, the character shape of the image data is corrected based on the outline information of the character and stored or displayed. be able to. Therefore, by recognizing only the font type without performing character recognition, it is possible to avoid erroneous recognition of characters.
JP 2002-27935 A JP-A-8-123904

  However, for example, in the character recognition method of Patent Document 1, even if a standard pattern is registered for each font, the feature amount for identifying the character is uncertain, so that the identification accuracy is not sufficient. In Patent Document 1, character line inclination, the number of loops, line width, character area, and the like are listed as feature amounts, but none of them is sufficient to specify a character. Further, if the feature amount is increased in order to improve the recognition rate, the recognition speed is lowered.

  In Cited Document 2, since characters are stored as contour information, characters can be acquired in the same shape as a paper document. However, since the acquired character information is not a character code, it is difficult to handle when reprocessing with word processing software or the like. .

  An object of this invention is to provide the character recognition apparatus, the character recognition method, and the character data for character recognition which can recognize a character with high precision and high speed in view of said problem.

  In view of the above problems, the present invention provides a character recognition device for recognizing characters from image data of an optically scanned document, a font determination unit for determining a font of the character, and a character size determination for determining the size of the character. Means, character data for recognizing characters stored in correspondence with the font and character size of the character, the font determined by the font determining means and the character size determined by the character size determining means And character identification means for determining the character code of the character with reference to the character data.

  According to the present invention, since character data for recognizing characters is prepared for each character size and font, characters can be recognized with high accuracy.

  Further, in one embodiment of the present invention, when character bitmap data is divided into a predetermined number of pixels, the character data indicates the number of black pixels that appear from the left, right, top, or ground direction of the character. It is recorded for each pixel column from the direction.

  According to the present invention, by using character data in which the number of black pixels appearing from each direction of a character is used, character recognition can be started without reading the entire character, and the character can be recognized at high speed.

  In one embodiment of the present invention, the character data is pixel character data having white pixel information or black pixel information for each pixel when the bitmap data of the character is divided into a predetermined number of pixels. Features.

  According to the present invention, since a character is recognized by determining whether it is a white pixel or a black pixel for each pixel, the character can be recognized with high accuracy. In addition, character recognition can be started without reading the entire character, and the character can be recognized at high speed.

  In one embodiment of the present invention, the character data has a dimension between a plurality of two pixels when character bitmap data is divided into a predetermined number of pixels.

  According to the present invention, since the character is recognized based on the dimension between the two pixels of the character, the character can be recognized with high accuracy.

  Further, according to one aspect of the present invention, there is provided an inclination angle determination unit that determines an inclination angle of the document, and the character identification unit inclines the pixel character data according to the inclination angle determined by the inclination determination unit, and It is characterized by recognition.

  According to the present invention, since the inclination angle of the original is determined and the pixel character data is also inclined, the character recognition can be started without reading the entire character, and the character can be recognized at high speed.

  In one embodiment of the present invention, there is provided character data creating means for creating the character data based on bitmap data of characters generated from the font data. The character data creating means includes a font discriminated by the font discriminating means. The character data is generated based on the bitmap data of the character determined by the character size determining means, and the character recognizing means recognizes the character based on the character data generated by the character data generating means. , And.

  According to the present invention, even a manuscript printed with a font that is used infrequently can recognize character of the font and create character data, so that the character can be recognized with high accuracy corresponding to various fonts. .

  It is possible to provide a character recognition device, a character recognition method, and character data for character recognition that can recognize characters with high accuracy and high speed.

  Hereinafter, the best mode for carrying out the present invention will be described with reference to the accompanying drawings. In addition, since the character recognition method of this invention is used for embodiment of the character recognition apparatus of this invention, embodiment of the character recognition method of this invention is described together in embodiment of a character recognition apparatus.

  FIG. 1 is an overall configuration diagram of a character recognition system including a character recognition device according to an embodiment of the present invention. In the character recognition system, a character recognition device 1, a scanner device 2, and a printer 3 are connected via a network 4 so that they can communicate with each other. The character recognition device 1 has character data 5 to be described later. The character recognition device according to the present embodiment may be integrated with the scanner device 2 or may be integrated with the scanner device 2 and the printer 3. Further, the scanner device 2 may have a facsimile function.

  An outline of a character recognition method of the character recognition device will be described. The character recognition device 1 has character data 5 in which bitmap data of each character is digitized for each font such as MS Mincho, MS Gothic, OSAKA. The character data 5 is stored for each character size (point).

  The bitmap data takes a black or white value for each pixel, but if the font and character size are specified, the bitmap data of the character is a binary (monochrome) image specific to the font and character size. Become. Therefore, by recognizing characters using such character data 5, characters can be recognized with extremely high accuracy.

  When the scanner device 2 reads a document on which characters created with these fonts are printed and transmits the document to the character recognition device 1, the character recognition device 1 determines the font type and character size of the printed characters, Character recognition is performed by referring to the character data 5 according to the determined font and character size.

  The scanner device 2 has a well-known configuration. For example, an original set on a contact glass is optically scanned, light is applied to the original, and the reflected light is input to a one-dimensional image pickup device such as a CCD to generate an electrical signal. Convert. The scanner device 2 converts the converted electrical signal into digital data by controlling the moving speed of the light source and the paper feed, and converts it into digital data. Shading processing, scaling processing, edge processing, γ correction, binary processing Image data is acquired by performing well-known image processing such as conversion processing. The scanner device 2 may be a flat bed type or a document feed type that sends a document to a fixed light source.

  The printer device 3 has a known configuration, and prints print data transmitted from the character recognition device 1 or the scanner device 2. The image forming method of the printer 3 may be any method such as a laser method, an LED method, a liquid crystal shutter method, and an ink jet method.

  FIG. 2 shows an example of a hardware configuration diagram of the character recognition device. The character recognition device 1 is configured as a computer, for example. The character recognition device 1 includes a CPU 11, an input / output device 12, a display device 13, a drive device 14, a main recording device 15, an auxiliary storage device 16, and a communication device 17 that are connected to each other via a bus B.

  The input / output device 12 processes various operation signals input from the keyboard and mouse operated by the user, and inputs / outputs image data transmitted from the scanner device 2 and print data input / output to the printer 3. Control. The display device 13 displays a GUI (Graphical User Interface) such as various windows and data necessary for operating the character recognition device 1 on the display. The communication device 17 is an interface for connecting the character recognition device 1 to a network, and includes, for example, a NIC (Network Interface Card) or a modem.

  A character recognition program for operating the character recognition device 1 is provided by a recording medium 18 such as a memory card or a CD-ROM, or downloaded through the network 4. The recording medium 18 is set in the drive device 14, and data and programs are installed from the recording medium 18 to the auxiliary storage device 16 via the drive device 14.

  The auxiliary storage device 16 is configured by a hard disk device or a storage element, and stores an OS (Operating System), character data, a program, and necessary files and the like. The CPU 11 loads a character recognition program from the auxiliary storage device 16, develops the program in the main storage device 15, and executes the program.

  FIG. 3 shows an example of a functional block diagram of the character recognition device 1. The character recognition device 1 includes a font determination unit 21, a character size determination unit 22, an inclination angle determination unit 23, and a character identification unit 24. The font discriminating means 21 discriminates the character font of the image data obtained by the scanner device 2. The character size determining means 22 determines the character size of the image data. The inclination angle determination means 23 determines the inclination angle of the document or image data with respect to the main scanning direction. The character identification unit 24 determines the character code of the character by referring to the character data 5 based on the font and the character size.

  The character recognition program is a program that causes the CPU 11 to function as the font determination unit 21, the character size determination unit 22, the inclination angle determination unit 23, and the character identification unit 24.

  The character data 5 will be described. FIG. 4 shows bit map data of characters “KAN”. In FIG. 4, as an example, an MS Mincho font is used and the character size is 10.5 points. Since it is bitmap data, the shape of a character can be displayed if white or black pixels are arranged for each pixel. When characters are displayed in this manner as a bitmap, characters can be identified by dividing the scanned original character into similar pixels and comparing each pixel for black or white pixels.

  In FIG. 4, the number of pixels is 256 × 256. However, the number of pixels may be coarsened if a document written in English is recognized, or displayed at a high resolution such as 1024 × 1024. Good. For the sake of explanation, in FIG. 4, the lower left vertex is the origin, the right direction is the X direction, the upper direction is the Y direction, and the reverse direction is the -X direction and -Y direction, respectively.

  The character data of this embodiment is such a bit map in which the number of black pixels appearing from the left direction, right direction, top direction, or ground direction of a character is digitized for each pixel column. For example, when the number “black” appears from the left in the X direction, the number of black pixels X41, X42, and X43 appears in the fourth pixel row. In addition, X51 appears in the fifth pixel column in addition to X41, X42, and X43. Since the positions at which black pixels appear are different for each character, if the number of black pixels that appear for each pixel column in the X direction is used as character data, it becomes identification information unique to the character. That is, the character data of the present embodiment is the number of black pixels (0, 0, 0, 3, 1, 1,...) Appearing for each pixel column. Therefore, the number of character data in this embodiment is 256 (in practice, slightly smaller as will be described later).

  The number of black pixels that appear may be counted in the Y direction, or in the reverse direction of the X direction and the reverse direction of the Y direction. In the −Y direction, three black pixels −Y41, −Y42, −Y43, and −Y44 appear in the fourth pixel in the character “KAN”. Further, -Y51 to -Y57 appear in the fifth pixel. Thus, four character data in the X direction, Y direction, -X direction, and -Y direction can be extracted for one character.

  With such character data, character recognition can be started before reading the entire character, regardless of which direction the character is scanned. In addition, after reading the entire character, the character can be recognized using character data in all four directions.

  By the way, when actually recognizing a character, it is not known where the character starts until a black pixel appears (in other words, where the first pixel is not found in 256 pixel columns), so the character data is a pixel in which a black pixel exists. To the number of black pixels appearing for each pixel column. That is, if it is “Chinese” (0, 0, 0, 3, 1, 1,...), (0, 0, 0) is omitted.

  Further, even if the character bitmap data is an outline font, the character bitmap data has a slightly different shape when the character size is different. For example, a 12-point character and a 10.5-point character are not quite similar. For example, one black pixel appears at the fourth pixel at 10.5 points, but two black pixels appear at the fourth pixel at 12 points. The positions at which the black pixels appear vary depending on the size of the character. Therefore, the character data of this embodiment stores character data for each character size. The resolution of the size of characters to be stored may be any size that is generally used. For example, 8, 9, 10, 10, 10.5, 11, 12, 14, 16, 18, 20 It is about a point.

  FIG. 5 shows an example of character data. FIG. 5 shows the number of black pixels that appear in each of the four directions in association with the font type and size. The character data includes character data of other sizes.

  In FIG. 5, the number of pixels that are inverted from white to black is counted, but the pixels that are inverted from black to white may be counted as character data.

  In addition, since the character recognition apparatus according to the present embodiment performs character recognition for each character size, character data can be constructed using the dimensions of characteristic portions of the characters. FIG. 6 is a diagram showing an example of how to measure the size of characters. For example, the size of the top and bottom of the left and right is an example of the size of the character, and this determines the size of the entire character.

  As shown in FIG. 6, the bitmap data of characters can be divided by a portion where black pixels are continuous. In the case of “Kan”, each picture of Sansui, the upper part of the bowl and the lower part of the bowl are continuous parts. Two pixels having the largest distance in each continuous portion are extracted, and the dimension between the two pixels is used as character data. In FIG. 6, the lengths 1 to 3 of the three water images indicate the dimension between the pixels having the largest distance. Similarly, Length 4 is the dimension between the pixels where the distance is the longest at the upper part of the ridge and Length 5 is the lower part of the cocoon.

  Further, the interval between the continuous portions can be character data. For example, the distance between the pixels at the end of each continuous portion. In FIG. 6, the dimensions between the first and second strokes of the three waters are dis1 and dis3, and the dimension between the second and third strokes is dis2 and the first and third strokes. The dimension between the end portions of the second and third strokes was expressed as dis4 and dis5, respectively. In FIG. 6, the interval between the right ends or the left ends of each continuous portion is extracted, but the interval between the right end portion and the left end portion may be extracted. Similarly, it is possible to extract the interval between each drawing of the three waters and the upper and lower parts of the basket. Alternatively, the circumscribed rectangle of each continuous portion may be calculated, and the diagonal line of the circumscribed rectangle may be used as character data.

  Also, two pixels can be extracted in the vertical direction and the horizontal direction, and the dimension between the two pixels can be used as character data. In the “Kan” bitmap data, there are three black pixels in the leftmost pixel column and two black pixels in the rightmost pixel column, and two pixels are extracted from each to extract the dimension between the pixels. . In FIG. 6, the uppermost pixels of the leftmost and rightmost are extracted and the dimension is LR1, and the lowermost pixels of the leftmost and rightmost are extracted and the dimension is LR2.

  Also, in the “Kan” bitmap data, there are four black pixels in the uppermost pixel column and six black pixels in the lowermost pixel column. Extract. In FIG. 6, the leftmost pixels of the top and bottom pixel columns are extracted and the distance is TB1, and the rightmost pixels of the top and bottom pixel columns are extracted and the distance is TB2.

  FIG. 7 shows an example of character data using the size of the characteristic part of the character. The font type and size are the same as in FIG. In addition, the dimensions of the top and bottom, the left and right, the continuous portions 1 to n (n = 5 in the case of “Han”), the dimensions between the pixels in the top and bottom directions and between the pixels in the left and right directions are stored.

  In the present embodiment, the unit of dimension can be a unit representing an absolute distance (for example, mm, cm, etc.) instead of the number of pixels. Since the character data of the present embodiment is stored for each character size, accurate character recognition is enabled by evaluating the dimension of the characteristic part of the character with an absolute value.

  Further, in addition to the dimensions as shown in FIG. 6 or 7, an angle may be used as a parameter representing character characteristics. In FIG. 6, a straight line connecting two pixels is obtained when extracting the dimensions. The angle of the straight line is obtained, and the angle formed by each straight line is used. For example, the angle formed by dis1 and dis3 and dis1 and Length5. In this way, character recognition can be performed with higher accuracy by using the corners formed in addition to the dimensions of the characteristic portions.

  Next, determination of the font of characters printed on a scanned document will be described. It is assumed that the character size has already been determined when determining the font.

  FIG. 8 shows bitmap data for each font of the characters “go”. In FIG. 8, as an example, fonts of MS Mincho, MS Gothic, and HG font are shown. As shown in FIG. 8, the line thickness, the ratio of black pixels, and the shape of the MS Mincho, MS Gothic, and HG fonts differ greatly. Therefore, the font discrimination means 21 can discriminate the font based on the thickness of the line.

  In the case of the thickness of the line and the ratio of the black pixels, MS Mincho <HG book <MS Gothic. In FIG. 8, the thicknesses of the four lines are shown for each font as the thickness of the line. If the thickness of the line of the scanned character is detected in several parts, the font type is determined based on the average. it can. Alternatively, the line thickness may be calculated for some characters and determined based on the average.

  In addition, when the ratio of black pixels is used, the circumscribed rectangle of the scanned character is detected, and the font can be determined based on the ratio of the black pixel with respect to the area of the circumscribed rectangle. Since the proportion of black pixels varies depending on the character, the font discrimination means 21 obtains the proportion of black pixels for, for example, one line of characters or one page of characters and discriminates the font according to the average.

  Further, the font discrimination means 21 discriminates the font based on, for example, a change in the thickness of one stroke line. In MS Gothic, the thickness of one stroke line hardly changes, and in HG font, it changes greatly. Therefore, for example, the font can be determined by calculating the change rate of the thickness of the line near the start point (for example, A1, B1, C1) and the end point (A2, B2, C2) of a certain stroke.

  The character data 5 has information indicating the thickness of the character, the ratio of black pixels, and the rate of change in the thickness of the line in association with the size of the character and the font. Easily distinguish fonts.

  A process in which the character recognition device 1 recognizes a character based on the above configuration will be described with reference to the flowchart of FIG. A document is set on the scanner device 2 and scanned in order, and image data is sequentially transmitted to the character recognition device 1.

  First, the first line of the document is scanned to determine the character size and font (S11). When it is detected that the line between which there is no character is detected and the scanning of the first line is completed, first, the character size determining means 22 determines the size of the character (S12). The character size may be determined in any way. For example, a circumscribed rectangle of each character is obtained and determined based on the size of the circumscribed rectangle. Even if the characters are the same size, the size of the circumscribed rectangle is different for each character, so the determination is made using the average of several characters. Since the document size is known, it may be determined by the size of the circumscribed rectangle relative to the document size, or one circumscribed rectangle may be determined by the number of pixels detected by the image sensor.

  Next, the font discrimination means 21 discriminates the font type (S13). As described above, if the size of each character is known, the font can be determined based on information such as the thickness of the line of the character data 5.

  Next, the character recognition device 1 determines the top of the document, that is, the direction of the characters (S14). Since it is not known in which direction the character is printed on the document set on the scanner device 2, the character recognition device 1 determines the top of the character based on the shape of the character. For example, the top and bottom of the character is determined based on the aspect ratio of the circumscribed rectangle of each character, the direction of the straight line portion, the direction of payment, and the like. Note that, for example, character recognition may be performed by well-known pattern matching using a standard pattern, and the direction in which the character can be recognized may be detected to determine the top and bottom.

  Next, the character identification means 24 performs character recognition using the character data 5 (S15). The character identification means 24 recognizes the character by referring to the character data based on the determined character size and font. Since the entire first line has already been scanned, the character recognition means 24 may use any of the character data in the four directions in FIG.

  Next, the character recognition device 1 determines whether or not the character recognition of all lines has been completed (S16), and if not completed (No in S16), scans the next line (S17) and performs character recognition (S17). S18).

  Since the character size and font have already been determined in character recognition on the first line, character recognition can be started before reading the entire line on the second and subsequent lines. That is, since the vertical direction of the character is known, character data in any of the X direction to -Y direction in FIG. 5 is extracted according to the vertical direction, and the character recognition candidates are narrowed down for each line in the main scanning direction. Can do. For example, when three black pixels are detected by scanning in one line where black pixels appear first, character candidates having three black pixels are extracted at the beginning of character data, and then five black pixels are detected by scanning. In such a case, candidates are narrowed down to characters having five black pixels in the second pixel column of the character data. Therefore, the character recognition according to the present embodiment can recognize a character without reading the entire character, and can speed up the character recognition process. When the character recognition for all the lines is completed, the processing in the flowchart of FIG. 9 is completed.

  Note that when character recognition is performed after the entire line is scanned, character recognition may be performed using character data in any direction from the X direction to the -Y direction in FIG. Also good.

  In FIG. 9, character recognition is performed using the character data of FIG. 5, but character recognition may be performed using the size of the characteristic part of the character as illustrated in FIG. 7. In this case, it is preferable to recognize the characters by applying the dimensions after scanning each whole line, but it is also possible to gradually narrow down the character candidates based on the dimensions that can be extracted in the scanned range. .

  According to the present embodiment, character data is stored in advance according to the font and size, and compared with characters of a document scanned for each pixel with sufficient resolution, so that characters can be recognized with extremely high accuracy. The character data stored in advance may be a font that is frequently used (for example, MS Mincho, MS Gothic, HG font, etc.) and a frequently used size (10.5 points, 12 points, etc.). The capacity does not become too large. In addition, after the character recognition on the first line is completed, the character recognition can be performed without reading the entire character, so that the recognition speed can be improved.

  In the present embodiment, a character recognition device capable of recognizing characters in the same manner as in the first embodiment even when tilted will be described. In this embodiment, the structure of character data is different from that of the first embodiment. The system configuration diagram and the functional blocks of the image recognition apparatus 1 are the same as those in the first embodiment.

  As shown in FIG. 4, the bitmap data of characters is data in which black or white is arranged in each 256 × 256 pixel (hereinafter referred to as pixel character data). Therefore, the recognition rate can be further improved by using the pixel character data as it is for character recognition. Specifically, the origin of the bitmap data is appropriately determined, and 1 (black pixel) or 0 (white pixel) bit is stored in each 256 × 256 pixel. The image recognition apparatus 1 of this embodiment has pixel character data for each font type and character size. Note that if the pixel character data used for character recognition is decompressed and used, and the pixel character data is compressed otherwise, the file capacity can be reduced.

  As shown in FIG. 10A, when the document is scanned with an inclination, characters are read with an inclination with respect to the main scanning direction of the one-dimensional image sensor. FIG. 10B shows an example of image data when the original is read with an inclination. If the character is tilted as shown in FIG. 10B, the position of the pixel where the black pixel appears is shifted. Therefore, in this embodiment, the inclination angle of the original is detected, and the pixel character data of the character data is inclined according to the inclination angle to recognize the character.

  FIG. 11 is a flowchart showing a process in which the character recognition device 1 recognizes a tilted character. In FIG. 11, the same steps as those in FIG. 9 are denoted by the same reference numerals.

  First, the image recognition apparatus 1 scans the first line (S11), and the inclination angle determination means 23 detects the inclination angle based on the image data for one line (S20). The inclination angle may be detected in any way. For example, assuming a base line connecting the ground (minimum part) of one line of characters, an angle θ between the base line and the main scanning direction is calculated. If the document insertion angle is detected, the tilt angle can be obtained from the angle formed by the insertion angle and the main scanning direction.

  Next, the character recognition device 1 corrects the inclination angle θ of the image data obtained by scanning the document, and determines the character size, font, and top and bottom as in the first embodiment (S12 to 14), and FIG. Alternatively, character recognition is performed based on the character data of FIG. The character in the first line may be recognized by determining whether it is a black pixel or a white pixel for each pixel based on the pixel character data.

  Next, the character identification means recognition 24 corrects the pixel character data to have the inclination angle θ inclined (S21). For example, in the case of pixel character data such as a solid line shown in FIG. 12, a virtual frame such as a dotted line is set by inclining the angle θ. By performing such processing on the pixel character data, even if the document is tilted, one of the virtual frames coincides with the main scanning direction, so that character recognition can be performed using the pixel character data.

  For example, if side A of the virtual frame coincides with the main scanning direction, the number of black pixels appearing for each column from side A can be counted to apply the same as in the first embodiment even if the document is inclined. It can be character data.

  The subsequent steps are the same as those in the flowchart in FIG. That is, in the second and subsequent lines, character recognition can be started before reading the entire line. Since the vertical direction of the character and the inclination angle θ are known, the character recognition candidates can be narrowed down for each line in the main scanning direction based on the pixel character data corrected with the inclination angle θ.

  Therefore, the character recognition according to the present embodiment can recognize a character without reading the entire character even if the document is tilted, and can speed up the character recognition process. When the character recognition of all the lines is completed, the process of the flowchart in FIG.

  According to the present embodiment, in addition to the effects of the first embodiment, even when the document is inclined, character data is prepared according to the inclination angle, and high-speed character recognition is possible. Note that if character recognition is performed after the entire line is read, the inclination of the image data of the line can be corrected. Therefore, even if the character data of the first embodiment is used or the pixel character data without inclination correction is used. Can be recognized.

  When pixel character data is used, it is possible to determine black or white for each pixel position rather than the number of black pixels appearing, so that character recognition can be performed more accurately.

In the first and second embodiments, it is assumed that the character recognition device has character data or pixel character data (hereinafter simply referred to as character data), but generates character data according to the recognized font and character size. May be.
FIG. 13 shows a system configuration diagram for creating character data. The system configuration diagram of FIG. 13 is realized by a character recognition device 1 configured by a computer.

  Usually, the personal computer of the character recognition apparatus 1 stores a plurality of font data 31 provided by the OS. The font data 31 stores outline font data and bitmap data in association with character codes.

  The rasterizer 32 expresses a character corresponding to a character code as a collection of small colored points. The rasterizer 32 generates bitmap data 34 corresponding to the character code, font, and size when the font and character size of the character of the predetermined character code are designated by the application software.

  When the bitmap data 34 is divided into predetermined pixels (for example, 256 × 256) as shown in FIG. 4, black or white pixels are arranged in each pixel. That is, the character data creation means 36 can generate pixel character data by dividing the bitmap data 34 into predetermined pixels and determining whether each pixel position is black or white. Further, the character data creating means 36 can generate character data indicating the number of black pixels appearing based on the bitmap data 34 divided into pixels or the pixel character data.

  FIG. 14 is a flowchart for character recognition based on character data created by the character data creation means. In FIG. 14, the same steps as those in FIG. 9 are denoted by the same reference numerals.

  First, the character size is determined based on the scanned image data of one line (S11, 12, S14), and character recognition is performed (S30). The character recognition in step S30 is based on known pattern matching using a standard pattern. As a result, characters are recognized with a recognition rate with a predetermined accuracy for the first line.

  Next, the font discrimination means 21 discriminates the font of the character on the first line (S31). Since the character in the first line has already been recognized, the rasterizer 32 generates character bitmap data in a plurality of fonts based on the character code of the recognized character. Then, pattern matching is performed again on the characters in the first line by using the bitmap data of each font. The font discriminating means 21 discriminates a font having a good matching among a plurality of fonts as a font used in the document.

  Since the font is determined, the character data creating means creates character data and pixel character data of the font and size (S32). As a result, character data is obtained in the same manner as in the first or second embodiment.

  The subsequent steps are the same as in the first embodiment, and characters are recognized using the created character data until all lines are completed (S16) (S17, S18). Since the character data or pixel character data indicates whether each pixel is a black pixel or a white pixel, the character can be recognized with extremely high accuracy.

  According to the present embodiment, even a manuscript printed with a font that is used infrequently can be used to create character data by recognizing the font, so that characters can be recognized with high accuracy corresponding to various fonts. it can.

1 is an overall configuration diagram of a character recognition system including a character recognition device. It is an example of the hardware block diagram of a character recognition apparatus. Example of functional block diagram of character recognition device This is bitmap data of the characters “Kan”. It is an example of character data. It is a figure which shows an example of how to take the dimension of a character. It is an example of the character data using the dimension of the characteristic part of a character. This is bitmap data for each font of the character “go”. It is a flowchart figure of the process which a character recognition apparatus recognizes a character. FIG. 6 is a diagram illustrating an example of a document scanned at an angle. It is a flowchart figure which shows the process which recognizes the inclined character. It is an example of pixel character data inclined at an inclination angle θ. It is a system configuration figure of a character data creation system. It is a flowchart figure in the case of character recognition by the character data created by the character data creation means.

Explanation of symbols

1 Character recognition device 2 Scanner device 3 Printer 4 Network 5 Character data

Claims (11)

In a character recognition device that recognizes characters from image data of an optically scanned document,
Font discrimination means for discriminating the font of the character;
Character size determining means for determining the size of the character;
Character data for recognizing the character stored in association with the font of the character and the size of the character;
Character identifying means for determining a character code of the character with reference to the character data based on the font determined by the font determining means and the size of the character determined by the character size determining means;
A character recognition device comprising: The character data is obtained by dividing character bitmap data into a predetermined number of pixels.
The number of black pixels appearing from the left direction, right direction, top direction or ground direction of the character is recorded for each pixel column from the direction,
The character recognition device according to claim 1. The character data is obtained by dividing character bitmap data into a predetermined number of pixels.
It is pixel character data having information on white pixels or black pixels for each pixel.
The character recognition apparatus according to claim 1. The character data is obtained by dividing character bitmap data into a predetermined number of pixels.
Having a dimension between a plurality of two pixels,
The character recognition device according to claim 1, wherein the character recognition device is a character recognition device. An inclination angle determining means for determining an inclination angle of the original;
The character identification means recognizes the character by inclining the pixel character data according to the inclination angle determined by the inclination determination means;
The character recognition device according to claim 3. Character data creation means for creating the character data based on bitmap data of characters generated from font data,
The character data creating means creates the character data based on the bitmap data of the character determined by the font determining means and having the size determined by the character size determining means,
The character recognition means recognizes the character based on the character data created by the character data creation means;
The character recognition device according to claim 1, wherein the character recognition device is a character recognition device. In character data for recognizing characters from image data of an optically scanned document,
In association with the font of the character and the size of the character,
When the bitmap data of a character is divided into a predetermined number of pixels, the number of black pixels appearing from the left direction, right direction, top direction, or ground direction of the character is recorded for each pixel column from the direction. ,
Character data characterized by that. In character data for recognizing characters from optically scanned original image data,
In association with the font of the character and the size of the character,
When character bitmap data is divided into a predetermined number of pixels,
It is pixel character data having information on white pixels or black pixels for each pixel.
Character data characterized by that. In character data for recognizing characters from image data of an optically scanned document,
In association with the font of the character and the size of the character,
When character bitmap data is divided into a predetermined number of pixels,
Having a dimension between a plurality of two pixels,
Character data characterized by that. In a character recognition method for recognizing characters from image data of an optically scanned document,
A font determining step for determining a font of the character;
A character size determining step for determining the size of the character;
Based on the font determined by the font determining step and the character size determined by the character size determining step, refer to the character data stored in association with the font of the character and the character size. A character identification step for determining a character code of the character;
A character recognition method characterized by comprising: The character data is obtained by dividing character bitmap data into a predetermined number of pixels.
Recording the number of black pixels appearing from the left direction, right direction, top direction or ground direction for each pixel column from the direction,
Pixel character data having white or black pixel information for each pixel;
Or a dimension between a plurality of two pixels,
The character recognition method according to claim 10, wherein:

Images