JP2006023987A - Character matrix generation device - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To obtain a plurality of character candidates corresponding to each character and generate an appropriate character matrix, even if a recognition object image is such an image that the character position thereof is hardly specified due to the difference of brightness (density) and contrast. <P>SOLUTION: A number of character candidates are obtained by scanning and verifying a first storage means 1 storing the recognition object image by a reference pattern of character, which is read out of a database 4, using a character candidate generating means 8, and repeating this with respect to each reference pattern. The large number of character candidates including candidate characters, center coordinates and sizes, which are resulted from character recognition, are generated. These are associated with the coordinate around the center of each character among the recognition object images inputted from an input means 5 on the basis of operation of a user by a character arrangement means 9, thereby generating the character matrix. <P>COPYRIGHT: (C)2006,JPO&NCIPI

Description

  The present invention performs character recognition processing on a recognition target image including characters, and among a plurality of obtained character candidates, a plurality of character candidates corresponding to the same character in the recognition target image are determined for each character. The present invention relates to a technique for generating a line-up character matrix.

  When recognizing the character from the recognition target image including the character, generally, the character position corresponding to each character in the recognition target image (Note that the character position here means a region having a certain extent or more). And the recognition process is performed on the image at each character position. At this time, as a recognition result, a plurality of character candidates of the first character compensation, the second character compensation, the third character compensation,... Was able to select any letter complement as the correct answer.

  On the other hand, in addition to the above-described recognition processing (character recognition processing in a narrow sense) for character position unit images, the character candidate ranking is obtained by performing analysis processing as a language on various combinations of obtained character candidates. Various proposals have been made to enable error detection / correction in character recognition, including changes to the above.

One of them is a character matrix (character candidate) in which a plurality of candidate characters corresponding to the same character in the recognition target image among a large number of candidate characters obtained by character recognition processing are arranged for each character. A high-precision character recognition by searching using a word model and a language model such as a word chain probability between character candidates corresponding to each character and a word chain probability including a word candidate. There is a patented invention in which results can be obtained (see Patent Document 1).
Japanese Patent No. 3309174 (character recognition method and apparatus)

  By the way, if the recognition target image including characters is an image obtained by capturing a document including printed characters with a scanner or the like, the character position described above is between the character portion and the other portion, that is, the ground portion. Identify from the difference in brightness (density) and contrast, for example, create a histogram that counts the number of pixels (usually black pixels) that compose characters in the horizontal direction for all images, identify rows, By specifying a column by creating a histogram in which the number of black pixels is counted in the vertical direction for each specified row image, the image can be specified relatively accurately. As a result, a plurality of character interpolations are obtained as recognition results for the image at each character position, and a character matrix to be input in the above-described patent invention can be generated, so that highly accurate character recognition can be realized.

  However, when the recognition target image including characters is an image obtained by photographing a landscape including a sign or a signboard with characters written with a digital camera or the like, the character portion and the background portion here. It is difficult to specify the character position accurately due to the difference in brightness (density) and contrast between them, and as a result, a large number of character candidates can be obtained from images at positions where no character originally exists. There is a problem that an appropriate character matrix cannot be generated.

  The present invention has been made in view of the above points, and is an image in which recognition target images including characters are difficult to specify character positions corresponding to individual characters due to differences in luminance (density) and contrast. However, it is an object of the present invention to provide a character matrix generation apparatus, a method and a program for obtaining a plurality of character candidates corresponding to each character and generating an appropriate character matrix.

  In order to achieve the above object, the character matrix generation device of the present invention performs character recognition processing on a recognition target image including characters, and selects the same character in the recognition target image from among the obtained many character candidates. A character matrix generation device that generates a character matrix in which a plurality of corresponding character candidates are arranged for each character, and stores a database that stores a large number of reference patterns of characters to be recognized and an input recognition target image A number of character candidates obtained by reading a reference pattern of characters from the first storage means and the database, scanning and collating the first storage means with the read reference pattern, and repeating this for each reference pattern A character supplement generation means for generating a large number of character supplements including at least the candidate characters of the character recognition result, the center coordinates of the supplement characters, and the size of the candidate characters Second storage means for storing the generated large number of character candidates; input means for inputting coordinates near the center of each character in the recognition target image based on a user operation; and first storage Display means for displaying the recognition target image stored in the means and the input coordinates by the input means, and third storage means for storing coordinates near the center of each character in the recognition target image input by the input means And reading the coordinates near the center of each character in the recognition target image input by the user from the third storage means, and reading a large number of character interpolations from the second storage means, and the center of each character in the recognition target image Character arrangement means for generating a character matrix by associating a large number of character candidates with nearby coordinates is provided.

  In the present invention, it is also possible to use a character supplement generation means that can give not only a supplementary character, the center coordinates of the supplementary character and the size thereof, but also a score indicating the character likeness to the character supplement. is there.

  In the character complement generating means, a large number of character candidates at arbitrary positions in the recognition target image are generated, so that there is little omission of correct answer compensation. However, since it is generated in large quantities, it is difficult to find only correct answers that are filled with errors.

  In the present invention, the user (operator) designates the approximate coordinate position of the center of each character in the recognition target image by the display means and the input means, and the number of characters is designated from the number of the designated coordinates. become. The designation of coordinates can be realized by mouse operation, pen operation, and key operation.

  The character arrangement means generates a character matrix in which a plurality of character candidates corresponding to the same character are arranged for each character by associating each character complement with the coordinates designated by the input means.

  If this character matrix is used as an input, it is possible to obtain a more accurate character recognition result according to the above-described patented invention.

  As described above, according to the present invention, if further obtained from the coordinates near the center of each character in the recognition target image input by the user and a large number of character candidates, the score of the character compensation is also taken into account. Thus, a character matrix can be generated, and as a result, a more accurate character recognition result can be obtained.

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

<First Embodiment>
FIG. 1 shows a character matrix generation apparatus according to a first embodiment of the present invention, in which 1 is a first storage means, 2 is a second storage means, 3 is a third storage means, 4 is a character pattern database, 5 is input means, 6 is display means, and 7 is a central processing unit (CPU).

  The first storage unit 1 stores a recognition target image that is read from a storage medium and input or input from another device (such as a digital camera) via a communication medium. The 2nd memory | storage means 2 memorize | stores many character candidates mentioned later. The third storage unit 3 stores coordinates near the center of each character in the recognition target image input by the input unit 5.

  The character pattern database 4 stores many reference patterns of characters to be recognized. The input means 5 is for inputting coordinates near the center of each character in the recognition target image based on a user operation, and a mouse, a digitizer, a keyboard, or the like is used. The display means 6 displays the recognition target image stored in the first storage means 1 and the input coordinates by the input means 5 (specifically, displays a cursor or the like indicating the input coordinate position with respect to the recognition target image). For example, a CRT or a liquid crystal display is used.

  The CPU 7 controls the above-described units in accordance with the program shown in the flowchart of FIG. 2 and configures the character candidate generation means 8 and the character arrangement means 9 at this time.

  The character complement generating means 8 is obtained by reading a character reference pattern from the character pattern database 4, scanning and collating the first storage means 1 with the read reference pattern, and repeating this for each reference pattern. A large number of character candidates including a candidate character as a result of character recognition, the center coordinates of the complement character, and the size of the candidate character are generated.

  In addition, the character arrangement unit 9 reads out the coordinates near the center of each character in the recognition target image input by the user from the third storage unit 3 and reads out many character candidates from the second storage unit 2 for recognition. A character matrix is generated by associating a number of character candidates with coordinates near the center of each character in the target image, and is output.

  Hereinafter, a method for generating a character matrix by the device of the present invention will be described.

  First, when a recognition target image is read from a storage medium and input or input from another device via a communication medium, the CPU 7 stores the image in the first storage unit 1 (s1). Next, the CPU 7 reads out the character reference pattern from the character pattern database 4 by the character interpolation generating means 8, scans and collates the first storage means 1 with the read reference pattern, and refers to this. A large number of character candidates obtained by repeating the pattern, including a candidate character of the character recognition result, the center coordinates of the complement character, and the size of the candidate character are generated (s2 These are stored in the second storage means 2 (s3).

  FIG. 3 shows an example of the character compensation generated by the character compensation generating means 8. One character candidate is a plurality of data separated by “,”, that is, a supplementary character (actually from the left). Is the corresponding character code), the x-coordinate of the center, the y-coordinate of the center, and the size.

  Next, the CPU 7 displays the recognition target image stored in the first storage unit 1 and the input coordinates by the input unit 5 on the display unit 6 (s4), and input from the input unit 5 based on the user's operation. The coordinates near the center of each character in the recognition target image are stored in the third storage means 3 (s5).

  FIG. 4 shows an example of coordinates near the center of each character in the recognition target image input by the user. One center coordinate is enclosed in parentheses and a plurality of data separated by “,”, that is, It consists of the center x coordinate and the center y coordinate. Here, there are four coordinates, and the recognition target character is four characters.

  When the input and storage of the coordinates near the center of all the characters in the recognition target image are completed (s6), the CPU 7 causes the character placement unit 9 to display the recognition target image input by the user from the third storage unit 3. The coordinates near the center of each character are read out, and a large number of character candidates are read out from the second storage means 2, and a character matrix is created by associating a large number of character candidates with the coordinates near the center of each character in the recognition target image. Generate and output (s7).

  There are various association methods in s7, and specific examples thereof will be shown below.

<Specific example 1>
Here, the case of associating a character complement that contains the coordinates designated by the user as an association method in s7 of FIG. 2 will be described.

One of the coordinates designated by the user is P = (X, Y). If the center coordinates of a certain letter complement C are (A, B) and the size of the letter complement is W,
A−W / 2 ≦ X ≦ A + W / 2
And B−W / 2 ≦ Y ≦ B + W / 2
, The character candidate C is associated with the coordinates of P. The above determination is performed for all the coordinates specified by the user, and a character matrix is generated.

  In this example, the character area is calculated at a magnification of half the size of the supplementary character. However, the character area is calculated using various magnifications in consideration of the spacing between characters. It is also possible to do this.

<Specific example 2>
Here, as an association method in s7 of FIG. 2, a case will be described in which each character candidate is associated with the closest coordinate among the coordinates designated by the user.

One of the coordinates designated by the user is P = (X, Y). Let the center coordinates of a certain letter complement C be (A, B),
| X-A | + | Y-B |
Or (| XA | ² + | YB | ² ) ^1/2
Defines the distance between P and C. For each character compensation C, the coordinate P that minimizes the distance is searched, and a character matrix is generated.

<Specific example 3>
Here, as an associating method in s7 of FIG. 2, at each point of the input coordinates, a character complement whose center exists only in the vicinity away from the coordinate by a certain allowable distance is associated with each coordinate. A case where a character matrix is generated will be described.

One of the coordinates designated by the user is P = (X, Y). If the center of a certain letter complement C is (A, B) and Z is an allowable distance,
X−Z ≦ A ≦ X + Z
And YZ ≦ B ≦ Y + Z
The character complement C satisfying the condition is associated with the coordinate P. The above determination is made for all character candidates C, P is searched, and a character matrix is generated.

<Second Embodiment>
FIG. 5 shows a character matrix generation apparatus according to a second embodiment of the present invention, in this case, a score indicating the likelihood (character likelihood) of a candidate character is also used as character candidate data in the first embodiment. The flowchart corresponding to this program is shown.

  That is, the CPU 7 in FIG. 1 reads the character reference pattern from the character pattern database 4 by the process of s11 in FIG. 5, scans and collates the first storage means 1 with the read reference pattern, A large number of character candidates obtained by repeating for each reference pattern, the candidate character of the character recognition result, the center coordinates of the complement character, the size of the candidate character, and the likelihood of the candidate character (character likelihood The character candidate generating means 8 for generating a large number of character interpolations including a score indicating), and in the recognition target image input by the user from the third storage means 3 by the process of s12 in FIG. The coordinates near the center of each character are read out, and a large number of character candidates are read out from the second storage means 2, and a large number of character candidates are matched to the coordinates near the center of each character in the recognition target image. And a character matrix by rearranging the ranking of a plurality of character candidates corresponding to the coordinates near the center of the same character using the score (specifically, the character candidate having a higher score is arranged higher). The character arrangement means 9 is generated and output. Other configurations and operations are the same as those in the first embodiment.

  FIG. 6 shows an example of the character compensation generated by the character compensation generating means 8 of the present embodiment. One character candidate is a plurality of data delimited by “,”, ie, compensation from the left. It is generated as data consisting of a character (actually a corresponding character code), a center x-coordinate, a center y-coordinate, a size, and a score indicating the likelihood (characteristic likelihood) of the candidate character.

  In this way, a plurality of character candidates associated with each character in the recognition target image are arranged in order from the highest score indicating the likelihood (characteristic character) of the complement character, so that the correct answer seems to be higher. It is possible to generate a character matrix in which character supplements are arranged.

<Third Embodiment>
In the second embodiment, a plurality of character candidates associated with the respective characters in the recognition target image are arranged in descending order of the character-likeness score. Calculates the average, median, or mode statistic for all characters in the character size, and scores higher in order from the closest to the value in multiple character interpolations associated with each character Or giving a high score in order from the closest to the statistic only for character interpolation having candidate character sizes within a certain range from the statistic, and for candidate characters outside the range It is also possible to generate a character matrix by discarding character candidates having a size.

  As a result, it is possible to avoid a character candidate having a significantly different size and to arrange a character candidate that is closer to the correct answer in a higher rank in the character matrix.

In the above-described patented invention, the probability P (X _ij | C _j ) that the character C _j is recognized as X _ij is approximated using the binomial distribution as shown in Equation (4) in Patent Document 1. By giving the probability value a value obtained by using the character compensation score obtained in the third embodiment as a probability, recognition based on an actual score becomes possible.

<Fourth embodiment>
FIG. 7 shows a character matrix generation apparatus according to the fourth embodiment of the present invention. In the figure, 11 is a first storage means, 12 is a second storage means, 13 is a third storage means, 14 is a character pattern database, 15 is an input means, 16 is a display means, and 17 is a central processing unit (CPU).

  The first storage unit 11 stores a recognition target image that is read from a storage medium and input or input from another device (such as a digital camera) via a communication medium. The second storage unit 12 stores coordinates near the center of each character in the recognition target image input by the input unit 5. The 3rd memory | storage means 13 memorize | stores many character candidates mentioned later.

  The character pattern database 14 stores many reference patterns of characters to be recognized. The input means 15 is for inputting coordinates near the center of each character in the recognition target image based on a user operation, and a mouse, a digitizer, a keyboard, or the like is used. The display unit 16 displays the recognition target image stored in the first storage unit 1 and the input coordinates by the input unit 5 (specifically, displays a cursor or the like indicating the input coordinate position with respect to the recognition target image). For example, a CRT or a liquid crystal display is used.

  The CPU 17 controls the above-described units in accordance with the program shown in the flowchart of FIG. 8 and configures the character candidate generation means 18 and the character arrangement means 19 at this time.

  The character complement generating unit 18 reads the character reference pattern from the character pattern database 4 and reads the coordinates near the center of each character in the recognition target image input by the user from the second storage unit 12. A number of character candidates obtained by scanning and collating near the center of each character in the recognition target image input by the user in the first storage means 11 with a pattern and repeating this for each reference pattern, A large number of character candidates including the candidate character of the recognition result, the center coordinates of the complement character, and the size of the candidate character are generated.

  In addition, the character arrangement unit 19 reads out the coordinates near the center of each character in the recognition target image input by the user from the second storage unit 12 and reads out many character candidates from the third storage unit 13 for recognition. A character matrix is generated by associating a number of character candidates with coordinates near the center of each character in the target image, and is output.

  Hereinafter, a method for generating a character matrix by the device of the present invention will be described.

  First, when the recognition target image is read and input from the storage medium or input from another device via the communication medium, the CPU 17 stores the image in the first storage unit 11 (s21).

  FIG. 9 shows an example of a recognition target image in which it is difficult to specify a character position. Here, a yellow and black striped pattern (however, yellow is displayed in white in the drawing) is used as a background and the character “CAUTIONAL caution” is displayed. It shows the case where it is written.

  Next, the CPU 17 displays the recognition target image stored in the first storage unit 11 and the input coordinates by the input unit 15 on the display unit 16 (s22), and is input from the input unit 15 based on the user's operation. The coordinates near the center of the character in the recognition target image are stored in the second storage unit 12 (s23).

  FIG. 10 shows an example of coordinates in the vicinity of the center of each character “Head”, “Up”, “Note”, and “Intention” of each character in the image to be recognized input by the user, here “Head Overhead” described above. One center coordinate is enclosed in parentheses and consists of a plurality of data separated by “,”, that is, a center x coordinate and a center y coordinate. Here, there are four coordinates, and the recognition target character is four characters.

  When the input and storage of the coordinates near the center of all the characters in the recognition target image are completed (s24), the CPU 17 reads out the character reference pattern from the character pattern database 14 by the character complement generating means 18, and The coordinates in the vicinity of the center of each character in the recognition target image input by the user are read from the storage unit 12 of the second, and each character in the recognition target image input by the user in the first storage unit 11 with the read reference pattern Is a large number of character candidates obtained by scanning and collating the vicinity of the center and repeating this for each reference pattern, including the candidate character of the character recognition result, the center coordinates of the complement character, and the size of the candidate character. Are generated (s25) and stored in the third storage means 13 (s26).

  FIG. 11 shows an example of the character compensation generated by the character compensation generating means 18 superimposed on the recognition target image of FIG. 8 together with a lead line from the black circle representing the center coordinates. As shown in the figure, a large number of character candidates having slightly different center coordinates are generated for one character.

  Finally, the CPU 17 reads out the coordinates near the center of each character in the recognition target image input by the user from the second storage unit 12 by using the character arrangement unit 19 and also outputs a number of character strings from the third storage unit 13. The complement is read out, a character matrix is generated by associating a large number of character candidates with coordinates near the center of each character in the recognition target image, and output (s27). As the association method in s27, the method as described in specific examples 1 to 3 can be used as in the case of the first embodiment.

  FIG. 12 shows an example of a character matrix generated by the character placement means 19 (however, only the candidate characters here). “Head”, which is a character in the recognition target image designated by the user from the left, A character complement column associated with “upper”, “note”, and “intention” is shown. Here, three character supplements are described in each column, but the number is not limited to three, and any number of character supplements can be included.

  By providing such a character matrix as the input of the above-described patented invention, the “overhead warning” sequence emerges as a higher-level supplement, thereby providing an effect of obtaining a more accurate character recognition result.

  Thus, only the character compensation near the coordinates designated by the user is generated by the character candidate generation means 18, and the obtained character candidates are associated with the coordinates designated by the user, so that unnecessary character compensation is performed. Generation of characters can be suppressed, and efficient character compensation can be generated.

<Fifth embodiment>
FIG. 13 is a character matrix generation apparatus according to a fifth embodiment of the present invention, here a score using a score indicating the likelihood (character likelihood) of a candidate character as character candidate data in the fourth embodiment. The flowchart corresponding to this program is shown.

  That is, the CPU 17 in FIG. 7 reads the reference pattern of characters from the character pattern database 4 by the process of s31 in FIG. 13, and near the center of each character in the recognition target image input by the user from the second storage unit 12. Is obtained by scanning and collating the vicinity of the center of each character in the recognition target image input by the user in the first storage means 11 with the read reference pattern, and repeating this for each reference pattern. A large number of character candidates, including a candidate character of a character recognition result, a center coordinate of the complement character, a size of the candidate character, and a score indicating the likelihood (character character) of the candidate character The character candidate generating means 18 for generating the character complement is configured, and the recognition input by the user from the second storage means 12 by the process of s32 in FIG. The coordinates near the center of each character in the elephant image are read out, and a large number of character candidates are read out from the third storage means 13, and a number of character candidates are associated with the coordinates near the center of each character in the recognition target image, A character matrix is generated by rearranging the ranks of a plurality of character candidates corresponding to coordinates near the center of the same character using the score (specifically, the character candidates with higher scores are arranged higher). The character arrangement means 19 to be output is configured. Other configurations and operations are the same as those in the fourth embodiment.

  In this way, by arranging a plurality of character candidates associated with each character in the recognition target image in descending order of the score indicating the likelihood (characteristic character) of the complement character, the correct answer is further increased. It is possible to generate a character matrix in which the character supplements are arranged.

The block diagram which shows the character matrix production | generation apparatus concerning the 1st Embodiment of this invention The flowchart corresponding to the program of the character matrix production | generation apparatus concerning 1st Embodiment Explanatory drawing which shows an example of the generated character compensation Explanatory drawing which shows an example of the coordinate near the center of each character in the recognition target image input by the user The flowchart corresponding to the program of the character matrix production | generation apparatus concerning 2nd Embodiment Explanatory drawing which shows an example of the generated character compensation The block diagram which shows the character matrix production | generation apparatus concerning the 4th Embodiment of this invention The flowchart corresponding to the program of the character matrix production | generation apparatus concerning 4th Embodiment Explanatory drawing which shows an example of the recognition target image for which specification of a character position is difficult Explanatory drawing which shows an example of the coordinate near the center of each character in the recognition target image input by the user An explanatory diagram showing an example of the generated character complement superimposed on the recognition target image together with a black circle representing the center coordinates thereof Explanatory drawing which shows an example of the generated character matrix The flowchart corresponding to the program of the character matrix production | generation apparatus concerning 5th Embodiment

Explanation of symbols

  1, 11: First storage means, 2, 12: Second storage means, 3, 13: Third storage means, 4, 14: Character pattern database, 5, 15: Input means, 6, 16: Display Means 7, 17: Central processing unit (CPU), 8, 18: Character candidate generation means, 9, 19: Character arrangement means.

Claims (12)

A character matrix in which character recognition processing is performed on a recognition target image including characters and a plurality of character candidates corresponding to the same character in the recognition target image among the obtained many character candidates are arranged for each character. A character matrix generator for generating
A database that stores many reference patterns of characters to be recognized;
First storage means for storing the input recognition target image;
A plurality of character candidates obtained by reading a character reference pattern from the database, scanning and collating the first storage means with the read reference pattern, and repeating this for each reference pattern, A character supplement generation means for generating a number of character supplements including at least a candidate character, a center coordinate of the supplementary character, and a size of the candidate character;
Second storage means for storing the generated large number of character candidates;
An input means for inputting coordinates near the center of each character in the recognition target image based on a user operation;
Display means for displaying the recognition target image stored in the first storage means and the input coordinates by the input means;
Third storage means for storing coordinates near the center of each character in the recognition target image input by the input means;
The coordinates near the center of each character in the recognition target image input by the user are read from the third storage means, and a large number of character candidates are read from the second storage means, and the coordinates near the center of each character in the recognition target image are read. A character matrix generation device comprising character arrangement means for generating a character matrix by associating a large number of character candidates with coordinates. A plurality of character candidates obtained by reading a character reference pattern from the database, scanning and collating the first storage means with the read reference pattern, and repeating this for each reference pattern, A character supplement generating means for generating a number of character supplements including at least a candidate character, a center coordinate of the supplementary character, a size of the candidate character, and a score indicating the likelihood of the candidate character;
The coordinates near the center of each character in the recognition target image input by the user are read from the third storage means, and a large number of character candidates are read from the second storage means, and the coordinates near the center of each character in the recognition target image are read. A plurality of character candidates associated with the coordinates, and character arrangement means for generating a character matrix by rearranging the rankings of a plurality of character candidates corresponding to coordinates near the center of the same character using the score. The character matrix generation device according to claim 1, wherein A character matrix in which character recognition processing is performed on a recognition target image including characters and a plurality of character candidates corresponding to the same character in the recognition target image among the obtained many character candidates are arranged for each character. A character matrix generator for generating
A database that stores many reference patterns of characters to be recognized;
First storage means for storing the input recognition target image;
An input means for inputting coordinates near the center of each character in the recognition target image based on a user operation;
Display means for displaying the recognition target image stored in the first storage means and the input coordinates by the input means;
Second storage means for storing coordinates near the center of each character in the recognition target image input by the input means;
The character reference pattern is read from the database and the coordinates near the center of each character in the recognition target image input by the user are read from the second storage means, and the user in the first storage means inputs the read reference pattern. A number of character candidates obtained by scanning and collating near the center of each character in the recognition target image and repeating this for each reference pattern, the candidate character of the character recognition result and the center of the complement character Character candidate generation means for generating a number of character interpolations including at least the coordinates and the size of the candidate character;
Third storage means for storing the generated large number of character candidates;
The coordinates near the center of each character in the recognition target image input by the user are read from the second storage means, and a large number of character interpolations are read from the third storage means, and the coordinates near the center of each character in the recognition target image are read. A character matrix generation device comprising character arrangement means for generating a character matrix by associating a large number of character candidates with coordinates. The character reference pattern is read from the database and the coordinates near the center of each character in the recognition target image input by the user are read from the second storage means, and the user in the first storage means inputs the read reference pattern. A number of character candidates obtained by scanning and collating near the center of each character in the recognition target image and repeating this for each reference pattern, the candidate character of the character recognition result and the center of the complement character Character candidate generation means for generating a number of character interpolations including at least coordinates, a size of the candidate character, and a score indicating the likelihood of the candidate character;
The coordinates near the center of each character in the recognition target image input by the user are read from the second storage means, and a large number of character interpolations are read from the third storage means, and the coordinates near the center of each character in the recognition target image are read. A character placement unit that associates a large number of character candidates with coordinates and generates a character matrix by rearranging the rankings of a plurality of character candidates corresponding to coordinates near the center of the same character using the score. The character matrix generation device according to claim 3. A character matrix in which character recognition processing is performed on a recognition target image including characters and a plurality of character candidates corresponding to the same character in the recognition target image among the obtained many character candidates are arranged for each character. A character matrix generation method for generating
Along with a database storing a large number of reference patterns of characters to be recognized, a computer having a first storage means, a second storage means, a third storage means, a display means and an input means,
The calculator
Storing the input recognition target image in the first storage means;
A plurality of character candidates obtained by reading a character reference pattern from the database, scanning and collating the first storage means with the read reference pattern, and repeating this for each reference pattern, Generating a number of character candidates including at least a candidate character, a center coordinate of the complement character, and a size of the candidate character;
Storing the generated multiple character supplements in a second storage means;
Displaying the recognition target image stored in the first storage means and the input coordinates by the input means on the display means;
Storing in the third storage means coordinates near the center of each character in the recognition target image input from the input means based on the user's operation;
The coordinates near the center of each character in the recognition target image input by the user are read from the third storage means, and a large number of character candidates are read from the second storage means, and the coordinates near the center of each character in the recognition target image are read. And a step of generating a character matrix by associating a large number of character candidates with coordinates. A plurality of character candidates obtained by reading a character reference pattern from the database, scanning and collating the first storage means with the read reference pattern, and repeating this for each reference pattern, Generating a large number of character supplements including at least a candidate character, a center coordinate of the complement character, a size of the candidate character, and a score indicating the likelihood of the candidate character;
The coordinates near the center of each character in the recognition target image input by the user are read from the third storage means, and a large number of character candidates are read from the second storage means, and the coordinates near the center of each character in the recognition target image are read. A step of associating a number of character candidates with coordinates and generating a character matrix by rearranging the rankings of a plurality of character candidates corresponding to coordinates near the center of the same character using the score. The character matrix generation method according to claim 5. A character matrix in which character recognition processing is performed on a recognition target image including characters and a plurality of character candidates corresponding to the same character in the recognition target image among the obtained many character candidates are arranged for each character. A character matrix generation method for generating
Along with a database storing a large number of reference patterns of characters to be recognized, a computer having a first storage means, a second storage means, a third storage means, a display means and an input means,
The calculator
Storing the input recognition target image in the first storage means;
Displaying the recognition target image stored in the first storage means and the input coordinates by the input means on the display means;
Storing the coordinates near the center of each character in the recognition target image input from the input means based on the user's operation in the second storage means;
The character reference pattern is read from the database and the coordinates near the center of each character in the recognition target image input by the user are read from the second storage means, and the user in the first storage means inputs the read reference pattern. A number of character candidates obtained by scanning and collating near the center of each character in the recognition target image and repeating this for each reference pattern, the candidate character of the character recognition result and the center of the complement character Generating a number of character interpolations including at least the coordinates and the size of the candidate character;
Storing the generated multiple character supplements in a third storage means;
The coordinates near the center of each character in the recognition target image input by the user are read from the second storage means, and a large number of character interpolations are read from the third storage means, and the coordinates near the center of each character in the recognition target image are read. And a step of generating a character matrix by associating a large number of character candidates with coordinates. The character reference pattern is read from the database and the coordinates near the center of each character in the recognition target image input by the user are read from the second storage means, and the user in the first storage means inputs the read reference pattern. A number of character candidates obtained by scanning and collating near the center of each character in the recognition target image and repeating this for each reference pattern, the candidate character of the character recognition result and the center of the complement character Generating a number of character interpolations including at least coordinates, a size of the candidate character, and a score indicating the likelihood of the candidate character;
The coordinates near the center of each character in the recognition target image input by the user are read from the second storage means, and a large number of character interpolations are read from the third storage means, and the coordinates near the center of each character in the recognition target image are read. A process of associating a number of character candidates with coordinates and generating a character matrix by rearranging the rankings of a plurality of character candidates corresponding to coordinates near the center of the same character using the score. Item 8. The character matrix generation method according to Item 7. A character matrix in which character recognition processing is performed on a recognition target image including characters and a plurality of character candidates corresponding to the same character in the recognition target image among the obtained many character candidates are arranged for each character. A character matrix generation program for generating
In addition to a database that stores a large number of reference patterns of characters to be recognized, a computer having a first storage means, a second storage means, a third storage means, a display means, and an input means,
Storing the input recognition target image in the first storage means;
A plurality of character candidates obtained by reading a character reference pattern from the database, scanning and collating the first storage means with the read reference pattern, and repeating this for each reference pattern, Generating a number of character candidates including at least a candidate character, a center coordinate of the complement character, and a size of the candidate character;
Storing the generated multiple character supplements in a second storage means;
Displaying the recognition target image stored in the first storage means and the input coordinates by the input means on the display means;
Storing in the third storage means coordinates near the center of each character in the recognition target image input from the input means based on the user's operation;
The coordinates near the center of each character in the recognition target image input by the user are read from the third storage means, and a large number of character candidates are read from the second storage means, and the coordinates near the center of each character in the recognition target image are read. And a step of generating a character matrix by associating a large number of character candidates with coordinates. A plurality of character candidates obtained by reading a character reference pattern from the database, scanning and collating the first storage means with the read reference pattern, and repeating this for each reference pattern, Generating a large number of character supplements including at least a candidate character, a center coordinate of the complement character, a size of the candidate character, and a score indicating the likelihood of the candidate character;
The coordinates near the center of each character in the recognition target image input by the user are read from the third storage means, and a large number of character candidates are read from the second storage means, and the coordinates near the center of each character in the recognition target image are read. A step of associating a plurality of character candidates with coordinates and generating a character matrix by rearranging the rankings of a plurality of character candidates corresponding to coordinates near the center of the same character using the score. The character matrix generation program according to claim 9. A character matrix in which character recognition processing is performed on a recognition target image including characters and a plurality of character candidates corresponding to the same character in the recognition target image are arranged for each character among the obtained many character candidates. A character matrix generation program for generating
In addition to a database storing a large number of reference patterns of characters to be recognized, a computer having a first storage means, a second storage means, a third storage means, a display means, and an input means,
Storing the input recognition target image in the first storage means;
Displaying the recognition target image stored in the first storage means and the input coordinates by the input means on the display means;
Storing the coordinates near the center of each character in the recognition target image input from the input means based on the user's operation in the second storage means;
The character reference pattern is read from the database, the coordinates near the center of each character in the recognition target image input by the user are read from the second storage means, and the user in the first storage means inputs the read reference pattern. A plurality of character candidates obtained by scanning and collating near the center of each character in the recognition target image and repeating this for each reference pattern, and the candidate character of the character recognition result and the center of the complement character Generating a number of character interpolations including at least the coordinates and the size of the candidate character;
Storing the generated multiple character supplements in a third storage means;
The coordinates near the center of each character in the recognition target image input by the user are read from the second storage means, and a large number of character interpolations are read from the third storage means, and the coordinates near the center of each character in the recognition target image are read out. And a step of generating a character matrix by associating a large number of character candidates with coordinates. The character reference pattern is read from the database and the coordinates near the center of each character in the recognition target image input by the user are read from the second storage means, and the user in the first storage means inputs the read reference pattern. A number of character candidates obtained by scanning and collating near the center of each character in the recognition target image and repeating this for each reference pattern, the candidate character of the character recognition result and the center of the complement character Generating a number of character interpolations including at least coordinates, a size of the candidate character, and a score indicating the likelihood of the candidate character;
The coordinates near the center of each character in the recognition target image input by the user are read from the second storage means, and a large number of character interpolations are read from the third storage means, and the coordinates near the center of each character in the recognition target image are read. A step of associating a number of character candidates with coordinates and generating a character matrix by rearranging the rankings of a plurality of character candidates corresponding to coordinates near the center of the same character using the score. The program for generating a character matrix according to claim 11.

Images