JPH1166236A - Method and device for character recognition and storage medium stored with character recognition program - Google Patents

Abstract

PROBLEM TO BE SOLVED: To skillfully recognize an object character, which has large handwriting deformation such as variations in the tilt of a character line by finding the degree of direction contribution of a character pattern and using a correlation value between the degrees of direction contribution between character lines. SOLUTION: A pre-processing means 100 normalizes a binarized character pattern as to a position and size of a character and stores the results in a frame memory 101, and an intersection detecting means 102 scans the stored character pattern from a specific position in a specific direction to detect the intersection with a character part. A connection length measuring means 104 measures the connection length of black pixels in the specific direction as to the black pixels of the intersecting character part, and a direction contribution degree calculating means 106 calculates the degree of direction contribution representing the distribution states of black pixels by directions using the measured connection length. A correlation calculating means 108 calculates the correlation value between the 1st degree of direction contribution and the 2nd degree of direction contribution which are calculated, and a character pattern recognizing means 110 recognizes the character pattern by using the calculated correlation value.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a character pattern recognizing method and a character recognizing device, and more particularly to a multi-character such as a handwritten kanji for a binarized character pattern obtained by optical conversion. In order to recognize a character object having a variety of handwritten deformations, a method and an apparatus for extracting a character line structure feature from a character pattern and recognizing an input character pattern, and a program for recognizing the input character pattern are stored. Related to a storage medium.

[0002]

2. Description of the Related Art As a conventional character pattern recognition method and apparatus, a character pattern subjected to binarization and normalization of position and size is observed from coordinate axes in a plurality of directions. The number of character lines crossing the character part in the direction orthogonal to the coordinate axis is counted, a feature vector pattern is created from the number of character lines, and matching with the previously stored feature dictionary table for each character is performed to recognize the character pattern. There is a method and apparatus for performing In the conventional character pattern recognition method and apparatus, it is possible to distinguish the difference in the general complexity of the character line structure caused by the difference in the character type according to the number of character lines traversed at the time of counting. However, since there is no information indicating a more detailed difference in the character line structure, many similar characters,
There is a problem that it is difficult to recognize a character object having various handwritten deformations with high accuracy.

[0003] A character pattern that has been subjected to binarization and position and size normalization is observed from coordinate axes in a plurality of directions. There is known a method and apparatus for recognizing a character pattern by obtaining the directional contribution of a character line described in Japanese Patent Application No. 56-46659. However, in the case of this type of character pattern recognition method and apparatus, there is a problem that a character object having many handwritten deformations such as a change in inclination of a character line cannot be recognized with high accuracy.

[0004]

SUMMARY OF THE INVENTION In view of the above-mentioned problems of the prior art, the present invention relates to a character pattern subjected to binarization and normalization of position and size, such as a change in inclination of a character line. It is an object of the present invention to provide a character recognition method and apparatus for skillfully recognizing a character object having many handwritten deformations such as variations in the inclination of a character line, and a storage medium storing a character recognition program. .

[0005]

In order to achieve the above-mentioned object, the present invention is characterized in that the character line is less susceptible to a change in the inclination of the character line and the like. In order to use the information, the directional contribution in the character pattern is obtained, and the correlation between the directional contributions between the character lines is calculated.

FIG. 1 is a diagram for explaining the principle of the present invention.
As shown in the figure, the character recognition method of the present invention scans a character pattern in a plurality of predetermined scanning directions in order to detect an intersection between a binarized character pattern and a character part ( Step 1) When the character portion intersects, the tentacle is extended in a plurality of predetermined directions for the black pixel of the intersecting character portion to determine the connection length of the black pixel for each direction (Step 2). From the connection length of the pixels, the directional contribution representing the distribution state of the character portion of the black pixel for each directional component is obtained (step 3). The value of the correlation between the directional contribution of the character portion at a certain crossing and the directional contribution of the character portion crossed before that time is calculated (step 4).
The character pattern is recognized using the correlation value (step 5).

According to the above method, even if the inclination of each character line varies, information on the parallelism of substantially the same character line can be obtained if the parallelism of the character lines is maintained. In addition, since it is less likely to be affected by the change in the inclination of the character line, it is possible to accurately recognize a character object having a large amount of handwritten deformation. FIG. 2 is a configuration diagram illustrating the principle of the character recognition device of the present invention. As shown in FIG. 1, the character recognition device 10 of the present invention includes a pre-processing unit 100, a frame memory 1
01, intersection detection means 102, connection length measurement means 104, direction contribution calculation means 106, correlation calculation means 108, character pattern recognition means 110, memory 112, and control means 11
Consists of four.

The preprocessing means 100 performs a normalization process on the position and size of the character in the binarized character pattern and stores the result in the frame memory 101. Intersection detection means 1
02 scans a character pattern pre-processed and stored in the frame memory 101 from a predetermined position in a predetermined direction,
The intersection with the character part is detected by scanning the character pattern.

[0009] The link length measuring means 104 measures the link length of the black pixel in the crossed character portion in a predetermined direction, and stores the measured link length in the memory 112. The directional contribution calculating means 106 calculates a directional contribution representing a black pixel distribution state in each of the plurality of directions for the black pixels of the intersecting character portion using the connection length stored in the memory 112, The directional contribution is stored in the memory 112.

The correlation calculation means 108 is provided in the memory 112
The correlation value between the first direction contribution and the second direction contribution stored in is calculated and stored in the memory 112. The character pattern recognition means 110 recognizes a character pattern using the value of the correlation stored in the memory 112. Control means 1
Reference numeral 14 controls the flow of processing for obtaining the directional contribution in the character pattern and calculating the correlation between the directional contributions between the character lines. Therefore, the control means 112 sequentially changes the predetermined position and the predetermined direction, activates the intersection detection means 102 for each position and each direction, and when an intersection is detected, The link length measuring means 104 is sequentially operated to operate the direction contribution calculating means 106 with the link lengths in a plurality of directions stored in the memory 112 for the black pixels of the intersecting character portion, and Counting the number of intersections with the character portion being scanned in the predetermined direction, and corresponding to the plurality of intersections, the directional contribution calculated at the plurality of intersections is defined as the first directional contribution, The other direction contribution calculated when the vehicle crosses before that is regarded as the second direction contribution, and the correlation calculation means 10 is used.
8 is operated to control the character pattern recognition unit 110 to operate after the correlation value is stored in the memory 112 for the predetermined position and the predetermined direction which are sequentially changed.

FIG. 3 is an operation flowchart of the character recognition apparatus 10 of the present invention. As shown in the figure, the character recognition device 10 of the present invention scans a binarized character pattern from a predetermined position in a predetermined direction (step 10), and the intersection of the character portion is performed by scanning the character pattern. It is determined whether or not there is any character (step 12). If the character intersects the character portion, the connection length of the black pixel in the crossed character portion is measured for a plurality of predetermined directions for the black pixel of the character portion. The directional contribution representing the distribution of the black pixels in the plurality of directions is calculated for the pixel (step 14), and it is determined whether the character portion intersected the character portion a plurality of times during scanning from the same position in the same direction (step 14). Step 16) In the case of a plurality of crossings, the correlation between one direction contribution calculated at the time of the plurality of crossings and another direction contribution calculated at the time of the previous crossing is calculated. Calculate (step 18) The predetermined position is sequentially changed, and the process returns to step 12 (step 20). The predetermined direction is sequentially changed, and the process returns to step 12 (step 22). The character pattern is calculated using the calculated correlation value. Is recognized (step 24).

[0012]

FIG. 4 is a block diagram for explaining an embodiment of a character recognition apparatus for realizing a character recognition method according to the present invention. As shown in FIG. 1, the character recognition device 10 of the present invention includes a preprocessing unit 12 to which an input character pattern is input, a feature extraction unit 14, and an identification processing unit 16 that generates an identification result.

The preprocessing unit 12 performs normalization processing on the binarized character pattern with respect to the position and size of the character.
The feature extraction unit 14 scans the character pattern obtained by the pre-processing unit 12 with the character pattern in a plurality of predetermined scanning directions. Extend the tentacles in a plurality of predetermined directions for the pixel to determine the connection length of black pixels for each direction,
A directional contribution degree representing a distribution state of each of the directional components of the character portion of the black pixel obtained from the black pixel connection length is obtained. Next, when the character extracting unit 14 intersects the character portion a plurality of times during scanning in the scanning direction, the feature extracting unit 14 determines the directional contribution of the character portion at a certain intersection among the plurality of intersections and the character intersecting before that. The correlation with the directional contribution of the part is obtained as a feature of the character pattern.

The identification processing section 16 performs a character pattern identification process using the above characteristics. Hereinafter, the normalization processing of the preprocessing unit 12 will be described in detail. A character pattern obtained by binarizing a character pattern obtained by photoelectric conversion or the like is given to the pre-processing unit 12 as an input character pattern, and the binarized character pattern is subjected to normalization processing for the position and size of the character. Do. The preprocessing unit 12 calculates the average value of the coordinate values of the x and y coordinates of all the black points constituting the input character pattern using, for example, a position normalization processing method known from the related art, and calculates the center of gravity of the character. And the parallel movement processing of the entire input pattern is performed so that the center of gravity of the character is aligned with the center position of the character frame. Also, the preprocessing unit 12 uses, for example, a normalization processing method of a conventionally known size so that the average value of the distance from the center of gravity to each writing point is equal to a predetermined normalization radius. Enlarge / reduce the character pattern uniformly around the center of gravity.

FIG. 5 is an explanatory diagram showing an example in which the input character pattern of the character "tree" is normalized by the normalization processing of the preprocessing unit 12. FIG. 7A shows an example of an input character pattern. FIG. 2B shows the character pattern after the preprocessing unit 12 has performed the position and size normalization processing on the input character pattern. In the above description, the preprocessing unit 12
When only the position and size normalization process is performed, but the character pattern after the position and size normalization process exceeds the character frame in the character "tree" as shown in FIG. In, it is possible to perform a frame removal process for removing a character portion that protrudes from a character frame. (D) of the same figure is an example in which the framing process is performed. Further, as shown in (e) of the figure, a smoothing process for filling or removing one mesh unevenness of a black point of a character line contour portion of the character pattern after the position and size normalization process is performed, respectively. Good.

The feature extraction unit 14, which is a main part of the present invention,
The character pattern normalized by the preprocessing unit 12 is input, and the character pattern is scanned in a plurality of predetermined scanning directions. For example, when the scanning direction is eight directions, the horizontal direction, the + 45 ° direction, and the +90
° direction, + 135 ° direction, + 180 ° direction, -135 °
The character pattern is scanned in the directions of -90 ° and -45 °. When the character intersects with the character portion, the tentacles are extended in a plurality of predetermined directions for the black pixels of the intersection character portion in the scanning direction. For example, in the case of 8 directions, 0 °, 45 °, 90 °, 135 °, 180 °,
The tentacles are extended in the directions of 225 °, 270 °, and 315 °, the number of connected black points is counted for each direction, and the process of calculating the directional contribution of the black points is performed. The directional contribution is described in Japanese Patent Application No. 56-46659. Furthermore,
When the character portion intersects the character portion M (M ≧ 2) times in the scanning in the above scanning direction, the directional contribution at the m (2 ≦ m ≦ M) intersection and the (m−1) th intersection A process for obtaining a correlation with the directional contribution is performed.

The identification processing unit 16 creates a feature table for identifying a character pattern based on the correlation value obtained by the feature extraction unit 14, and converts the feature table into an original using a conventionally known matching method. Then, the character pattern is identified by matching the stored character dictionary with the feature dictionary table. As a specific embodiment of the feature extracting unit 14, consider a case where a character is scanned in eight scanning directions. FIG. 6 is a diagram illustrating a vertical direction for observing the directional contribution of the character portion in the feature extraction unit 14 and coordinate axes orthogonal to the vertical direction. The eight directions are the horizontal direction, the + 45 ° direction, the + 90 ° direction, the + 135 ° direction, the + 180 ° direction, the −135 ° direction, the −90 ° direction, based on the horizontal direction.
And eight directions of -45 °, one in each direction,
The numbers 2, 3, 4, 5, 6, 7, 8 are given.

FIG. 7 shows a case where a character pattern is scanned twice in a vertical direction and crosses a character portion twice. The points indicated by numbers 1 and 2 in FIG. 3 represent black points at the points where the white and black points have changed from white points to black points when the character part is first and second times crossed by scanning. When a character portion intersects with such a scan, the black pixels in the intersecting character portion are scanned in eight directions (0 °, 45 °, 90 °, 135 °, 1
Extend the tentacles to 80 °, 225 °, 270 ° and 315 °). Here, in the 8 directions where the tentacles are extended,
Numbers 3, 4, 5, 6, 7, 8 are assigned.

FIG. 8A shows eight directions in which the tentacles are extended in order to determine the black dot connection length of the character pattern, and FIG.
FIG. 3 shows how to extend the tentacle from the black point of the character portion to obtain the black point connection length in order to obtain the directional contribution degree of the black point. With respect to the eight directions in which the tentacles are extended, the directional contribution of the black point is obtained, and the character pattern is identified.

Hereinafter, a first embodiment of the present invention will be described. The N × N mesh character pattern obtained by the preprocessing unit 12 is observed from eight directions with reference to the horizontal direction. In the s scanning direction (s = 1, 2,..., 8),
Scanning is performed in the s scanning direction passing through the position t on the coordinate axis and orthogonal to the coordinate axis. As shown in FIG. 6, the position t on the coordinate axes is t = 1, 2,. . . ,
N and t = 1, when s = 2,4,6,8
2,. . . , N ′, N ′ = 2N.

At the start of scanning, if it is assumed that the immediately preceding pixel is a white point, and if the character portion intersects the character portion m times in scanning in the s direction, the directional contribution A of the black point changed from a white point to a black point at the time of the intersection. _{m is, A m = (α 1,} α 2, α 3, α 4, α 5, α 6, α 7,
α ₈ ) _m is represented by an eight-dimensional vector. Where α ₁ , α ₂ ,
α ₃ , α ₄ , α ₅ , α ₆ , α ₇ , and α ₈ are directional contribution components in eight directions, respectively, and a black point connection length l _i ( i = 1
2,. . . , 8), for example,

[0022]

(Equation 1)

## EQU2 ## In this example, the Euclidean distance is applied as α _i , but a distance other than the Euclidean distance may be applied. Further, when the cross character portion and M (M ≧ 2) times by scanning, m (2 ≦ m ≦ M ) th intersecting at the direction contribution A _m of (m-1) th direction contribution at cross The correlation Γ _{m−1, m} between the degree A _m−1 and

[0024]

(Equation 2)

Or

[0026]

(Equation 3)

Is represented by the following equation. Of the correlations Γ _{m−1, m} thus obtained, m = 2 to m = m ′ (2 ≦ m ′)
By selecting a range up to m ′ ≦ M, the feature pattern f _st obtained by scanning from a position t on a coordinate axis orthogonal to the s-direction scanning is f _st = (Γ ₁₂ , Γ ₂₃ ,. Γ _{m'-1 m '} ). Therefore, the feature vector F of the character pattern
Is: F = (f ₁₁ , f ₁₂ ,..., F _1N , f ₂₁ ,
f ₂₂ ,. . . , F _{2N ′} , f ₃₁ , f _{32 ,.} . . , F _3N , f
₄₁ , f42 _,. . . , F _{4N ′} , f ₅₁ , f _{52 ,.} . . ,
_{f 5N, f 61, f 62} ,. . . , F _{6N '} , f ₇₁ ,
f _72,. . . , F _7N , f ₈₁ , f _{82 ,.} . . , F _{8N ′} ).

A characteristic table is created as a character pattern characteristic using a value obtained by grouping a plurality of elements of the characteristic vector F of the character pattern represented in this manner. The character pattern is identified by determining the identification function D (F) such as the Euclidean distance as a function. Here, the discrimination function calculates the distance value between the feature vector of the input character pattern and the feature vector for each character type in the previously stored feature dictionary table, and determines the smallest value of the distance value (depending on the function). (Which may be the largest value) are output as candidate characters.

The feature vector of the input character pattern is represented by F =
(F ₁ , f ₂ ,..., F _k ,..., F _K ), each character C _{i in the} feature dictionary table (1 ≦ i ≦ L, L is the total number of character types)
The feature vector of S _i = (s _i1 , s _i2 , ..., s _iK )
Then, for example, in the case of the Euclidean distance, C
_i (i = 1 to L)

[0030]

(Equation 4)

Is calculated, and the smallest value C _{i is obtained.}
Is output as the correct character pattern. Next, a second embodiment of the present invention will be described as another method. The N × N mesh character pattern obtained by the preprocessing unit 12 is observed from eight directions with reference to the horizontal direction. In the s scanning direction (s = 1, 2,..., 8), scanning is performed in the s scanning direction passing through the position t on the coordinate axis and orthogonal to the coordinate axis. The position t on the coordinate axis is, as shown in FIG.
When s = 1, 3, 5, 7, t = 1, 2,. . . , N, and for s = 2, 4, 6, 8, t = 1, 2,. . . ,
N ′ and N ′ = 2N.

At the start of scanning, assuming that the immediately preceding pixel is a white point, if the character portion intersects the character portion m times in scanning in the s direction, the directional contribution B of the black point changed from a white point to a black point at the time of the intersection. _{m is, B m = (β 1,} β 2, β 3, β 4) represented by _m becomes four-dimensional vectors. Where β ₁ , β ₂ ,
β ₃ and β ₄ are directional contribution components in four directions, respectively.
The tentacles are extended in eight directions from the black point, and using the black point connection length l _i (i = 1, 2,..., 8) obtained for each direction, for example,

[0033]

(Equation 5)

Is represented by In this example, the city block distance is applied as β _i , but a distance other than the city block distance may be applied. Further, when the character portion crosses the character portion M (M ≧ 2) times by scanning, m (2 ≦ m ≦ M)
Correlation Δ _{m-1, m} between the direction contribution B _m-1 times th intersecting at the direction contribution B _m of (m-1) th time crossing

[0035]

(Equation 6)

Or

[0037]

(Equation 7)

Is represented by the following equation. Of the correlations Δm _{−1, m} thus obtained, m = 2 to m = m ′ (2 ≦ m ′)
By selecting a range up to m ′ ≦ M, the characteristic pattern g _st obtained by scanning from a position t on a coordinate axis orthogonal to the s direction scanning is g _st = (Δ ₁₂ , Δ ₂₃ ,. Δ _{m'-1 m '} ). Therefore, the feature vector G of the character pattern
G = (g ₁₁ , g ₁₂ ,..., G _1N , g ₂₁ ,
g ₂₂ ,. . . , G _{2N ′} , g ₃₁ , g _{32 ,.} . . , G _3N , g
₄₁ , g42 _,. . . , G _{4N ′} , g ₅₁ , g _{52 ,.} . . ,
_{g 5N, g 61, g 62} ,. . . , G _{6N '} , g ₇₁ ,
g _72,. . . , G _7N , g ₈₁ , g _{82 ,.} . . , G _{8N ′} ).

A characteristic table is created as a characteristic of a character pattern using a value obtained by grouping a plurality of elements of the characteristic vector G of the character pattern represented in this manner. The character pattern is identified by determining the identification function D (G) such as the Euclidean distance as a function.

[0040]

DESCRIPTION OF THE PREFERRED EMBODIMENTS Hereinafter, an example in which the character recognition method of the present invention is applied will be described. FIG. 9 is a diagram showing an example in which the character recognition method according to the present invention is applied to a character pattern in which the inclination of a character line has changed due to handwriting deformation. Two points indicated by numbers 1 and 2 in FIG. 3A are the first and second intersections obtained by scanning the handwritten character from the left side to the right side of the character frame. It corresponds to the black point when changing from a point to a black point. Two points indicated by numbers 1 and 2 in (b) of the same figure are the first and second white spots obtained by scanning another handwritten character from the left to the right of the character frame. Corresponds to the black point when changing from the black point to the black point.

When the method according to the second embodiment of the basic invention is applied to points 1, 2, 3 and 4, m = 1 to m =
Tables 1 and 2 below show the values of B _a1 and B _a2 and B _b1 and B _b2 of the directional contributions obtained in the range of two character intersections.
Shown in

[0042]

[Table 1]

[0043]

[Table 2]

The correlation of the two directions contribution delta _a12 and delta _b12
Are shown in Tables 3 and 4 below, respectively.

[0045]

[Table 3]

[0046]

[Table 4]

According to Tables 1 and 2, when the directional contribution B is used as the feature vector, the value of the directional contribution changes under the influence of the inclination change. On the other hand, as can be seen from Tables 3 and 4, by using the correlation Δ as the feature vector, the tilt variation occurs while the balance of the characters is maintained, as shown in FIGS. 9A and 9B. In the case of a character pattern, the feature vector is less susceptible to tilt variations.

By repeating the above processing in the same manner as described above and using the correlation Δ obtained from the entire character pattern as a feature vector, it is possible to reduce the inclination of the character pattern while maintaining the parallelism of the character. Can correctly recognize characters. In the above embodiment, the description has been given based on the configuration requirements of the character recognition device. However, the present invention is not limited to this example. It is also possible to store and store it and install it in each computer constituting the character recognition device as needed to perform character recognition. Further, the constructed program is stored in a portable storage medium such as a floppy disk or a CD-ROM,
Such a system can be used for general purposes.

The present invention is not limited to the above embodiment, but can be variously modified and applied within the scope of the claims.

[0050]

As described above, according to the present invention,
Since the information on the relative arrangement of the character lines can be extracted by calculating the parallelism of the character lines, it is less likely to be affected by fluctuations in the inclination of the character lines. Will be possible.

[Brief description of the drawings]

FIG. 1 is a diagram illustrating the principle of the present invention.

FIG. 2 is a principle configuration diagram of the present invention.

FIG. 3 is an operation flowchart of the character recognition device of the present invention.

FIG. 4 is a configuration diagram of a character recognition device of the present invention.

FIG. 5 is an explanatory diagram of preprocessing.

FIG. 6 is an explanatory diagram of a scanning direction and a coordinate axis for observing a direction contribution.

FIG. 7 is an explanatory diagram of detection of intersection with a character portion by scanning.

FIG. 8 is an explanatory diagram of measurement of a black dot connection length.

FIG. 9 is a diagram showing a character pattern with handwritten deformation.

[Explanation of symbols]

 REFERENCE SIGNS LIST 10 character recognition device 100 preprocessing means 101 frame memory 102 intersection detection means 104 connection length measurement means 106 direction contribution calculation means 108 correlation calculation means 110 character pattern recognition means 112 memory 114 control means

Claims (6)

[Claims] 1. A character pattern is scanned in a plurality of predetermined scanning directions in order to detect a crossing of a binarized character pattern with a character portion. The tentacles are extended in a plurality of predetermined directions with respect to the black pixels of the intersecting character portion to determine the connection length of the black pixels in each direction, and from the connection length of the black pixels, the distribution of the character portion of the black pixel for each direction component A direction contribution representing a situation is obtained, and when a character portion intersects a plurality of times during scanning in the scanning direction, the direction contribution of a character portion at a certain intersection out of the plurality of intersections intersects before that. A character recognition method characterized by performing a process of recognizing a character pattern by using a value of a correlation with a direction contribution of a character portion. 2. A crossing detection step of scanning the binarized character pattern from a predetermined position in a predetermined direction to determine whether there is a crossing with a character portion by scanning the character pattern; When intersected, measure the connection length of black pixels in a plurality of predetermined directions for the black pixels of the intersected character portion,
A directional contribution calculating step of calculating a directional contribution representing a distribution state of the black pixels in the plurality of directions with respect to the black pixels of the intersecting character portion, and a character portion and a plurality of times during scanning from the same position in the same direction. When crossing, a correlation calculation step of calculating a correlation between one direction contribution calculated at the time of crossing the plurality of times and another direction contribution calculated at the time of previous crossing, By sequentially changing the predetermined position and the predetermined direction,
A scanning step of repeatedly performing the intersection detection step, the directional contribution calculation step, and the correlation calculation step, and a recognition step of recognizing a character pattern using the calculated correlation value. Character recognition method. 3. A pre-processing means for performing a normalization process on a position and a size of a character in a binarized character pattern, and a character pattern obtained by the pre-processing means
A character pattern is scanned in a plurality of predetermined scanning directions, and when the character pattern intersects, the tentacles are extended in a plurality of predetermined directions for black pixels of the crossed character part, and black pixels are connected in each direction. The length is obtained, the directional contribution representing the distribution state of each directional component of the character portion of the black pixel obtained from the black pixel connection length is obtained. A feature extraction unit that obtains, as a feature of a character pattern, a correlation between a direction contribution of a character portion at a certain intersection and a direction contribution of a character portion that intersected before that of the plurality of intersections; A character recognizing device comprising: an identification unit for performing a character pattern identification process. 4. A pre-processing means for normalizing the position and size of a character to a binarized character pattern, scanning the pre-processed character pattern from a predetermined position in a predetermined direction, Intersection detection means for detecting an intersection with a character portion by scanning the pattern; measuring a connection length of the black pixel in the crossed character portion in a predetermined direction, and storing the measured connection length in a memory. Measuring means, for the black pixels of the intersecting character portion, calculating a directional contribution representing a distribution state of the black pixels for each of the plurality of directions using a connection length stored in the memory, and calculating the directional contribution; Directional contribution calculating means for storing in the memory; correlation calculating means for calculating a correlation value between the first directional contribution and the second directional contribution stored in the memory; and storing in the memory. On A character pattern recognition means for recognizing a character pattern using a correlation value, and the predetermined position and the predetermined direction are sequentially changed, and the intersection detection means is operated for each position and each direction to detect a cross. In this case, the connection length measuring means is sequentially operated in a plurality of directions, and the directional contribution calculating means is operated by the connection lengths in the plurality of directions stored in the memory for the black pixels of the intersecting character portion. Activating, counting the number of intersections with the character portion being scanned in the predetermined position and in the predetermined direction, and corresponding to the plurality of intersections, calculating the directional contribution calculated at the plurality of intersections in the first direction. And the other direction contribution calculated at the time of intersection before the second direction contribution is used as the second direction contribution to operate the correlation calculation means, and the predetermined position and the predetermined position sequentially changed About the direction After the value of the serial correlation is stored in the memory, the character recognition device including a control means for controlling to actuate the character pattern recognition means. 5. A pre-processing process for normalizing a binarized character pattern with respect to the position and size of a character, and a plurality of predetermined scans performed on the character pattern obtained by the pre-processing process. The character pattern is scanned in the direction, and when it intersects with the character portion, the tentacles are extended in a plurality of predetermined directions for the black pixels of the intersected character portion to determine the connection length of the black pixel in each direction, Determine the directional contribution representing the distribution of the directional component of the character portion of the black pixel obtained from the connection length, and if the character portion intersects the character portion multiple times during scanning in the scanning direction, A feature extraction process of obtaining a correlation between a direction contribution of a character part at a certain intersection and a direction contribution of a character part intersecting before that as a characteristic of a character pattern, and a character pattern identification process using the characteristic. Cormorants storage medium storing a character recognition program, characterized in that it comprises an identification process. 6. A pre-processing process for normalizing a character position and a size of a binarized character pattern with respect to a position and a size of the character, scanning the pre-processed character pattern from a predetermined position in a predetermined direction, An intersection detection process of detecting an intersection with a character portion by scanning a pattern; measuring a connection length of a black pixel in a predetermined direction for a black pixel of the crossed character portion; and storing the measured connection length in a memory. A measurement process, for the black pixels in the intersecting character portion, calculate a directional contribution representing a distribution state of the black pixels in each of the plurality of directions using the connection length stored in the memory; A directional contribution calculation process stored in a memory, a correlation calculation process of calculating a correlation value between the first directional contribution and the second directional contribution stored in the memory, and storing the correlation value in the memory; The character pattern recognition process for recognizing a character pattern using the correlation value stored in the memory, and the predetermined position and the predetermined direction are sequentially changed, and the intersection detection process is performed for each position and each direction. When the intersection is detected, the connection length measurement process is sequentially performed in a plurality of directions, and the black pixels of the crossed character portion are connected with the connection lengths in the plurality of directions stored in the memory. Activating the direction contribution calculation process, counting the number of intersections with the character portion being scanned at a predetermined position and in a predetermined direction, and corresponding to the plurality of intersections, the direction calculated at the time of the plurality of intersections The correlation calculation process is activated by setting the contribution as the first directional contribution and the other directional contribution calculated at the time of the previous intersection as the second directional contribution. And storing a character recognition device program comprising a control process for controlling the character pattern recognition process to operate after the correlation value is stored in a memory for the predetermined position and the predetermined direction. Storage media.