JPH03246779A - Character recognizing device - Google Patents

Abstract

PURPOSE:To perform character recognition at high speed and with superior operability by dividing a prescribed area into arbitrary plural areas, and performing the feature extraction of a character pattern based on a result in which the character direction of each divided area is decided from an extracted sub pattern. CONSTITUTION:The prescribed area including the character string of a document, etc., is divided into the arbitrary plural areas with an area segmenting means 4, and a vertical sub pattern and a horizontal sub pattern area extracted from each divided area with a sub pattern extraction means 10, and the features of the vertical sub pattern and the horizontal sub pattern are extracted with a character direction decision means 20. The character direction of the character string in the prescribed area is decided based on the feature, and the feature extraction of a character pattern is performed based on a decided result. In such a way, it is possible to recognize the characters of the character strings on various kinds of documents, etc., at high speed and with superior operability.

Description

DETAILED DESCRIPTION OF THE INVENTION (Field of Industrial Application) The present invention is an optical character reading device that optically reads printed documents, etc.
The present invention relates to a high-speed, small-sized, and easy-to-operate character recognition device such as OCR (hereinafter referred to as OCR).

(Prior art) Conventionally, as a technology in this field, Japanese Patent Application Laid-Open No. 62-1
There was one described in Publication No. 66479.

In recent years, with the spread of Japanese information processing systems, the demand for inputting Japanese printed documents into computers has rapidly increased, and the need for printed kanji OCR has rapidly increased to speed up input and save labor. .

This printed kanji OCR is equipped with a character recognition device that reads kanji, etc. in a printed document, and
It is necessary to recognize the character patterns of the read kanji and other characters.

By the way, printed documents include horizontally written documents and vertically written documents such as newspapers and paperback books.

For example, if we take a business card as an example of a printed document, as shown in Figures 2 (a) to (d), horizontal business cards (Figure 2 (
a)) and a vertical business card (Fig. 2(b)). Second
Figure (C) is a pattern for a horizontally written business card (Figure 2 (a)), and Figure 2 (d) is a pattern for a vertically written business card (Figure 2 (b)).
This is an example of a pattern obtained by rotating the pattern in FIG. 2(C) by 90° counterclockwise.

In a conventional character recognition device for printed kanji OCR, an input pattern read by an image reading means is stored in an image storage means, and a character pattern is cut out one character at a time by a character cutting means from the input pattern stored in the image storage means. put out. Then, the features of the cut out character pattern are extracted by the feature extracting means, and the extracted features are compared with the contents of the dictionary storage means storing the matching dictionary mask by the matching means, and the characters of the input pattern are Recognition is now possible.

With such a character recognition device, when recognizing a form with the same shape and horizontal and vertical text directions as shown in Figure 2, or a document with a mixture of horizontal and vertical lines,
Correct character recognition could not be performed unless the operator, while looking at the form, designated horizontal or vertical writing for each form or character string using a keyboard or the like.

Therefore, in order to solve such a problem, the technique of the above-mentioned document uses four
A dictionary of types is stored in a dictionary storage means in advance. Then, the input pattern is recognized with the four types of dictionaries, and from the recognition results obtained by each dictionary, the direction of the dictionary that outputs the result with the highest degree of similarity is determined to be the character direction,
Characters were recognized based on the determination results.

(Problems to be Solved by the Invention) However, the apparatus with the above configuration has the following problems.

In the technique of the above-mentioned document, the character direction is determined and character recognition is performed based on the determination result. Since it is necessary to perform four types of identification, character recognition processing takes time. Moreover, 0°, 90°
Since four types of dictionaries such as , 180° and 270° must be prepared, the memory capacity of the dictionary storage means for storing them becomes large, resulting in the problem of an increase in the size of the device. Furthermore, although the technology in the above document can automatically recognize business cards written vertically or horizontally,
It was not possible to recognize printed documents other than business cards, and therefore the device had low versatility and was not technically satisfactory.

The present invention solves a problem that the prior art had, and for printed documents other than business cards, the operator must distinguish and specify whether the character direction is horizontal or dark blue.
The object of the present invention is to provide a character recognition device that solves problems such as complicated operator operations and long processing times, and an increase in the size of the device.

(Means for Solving the Problem) In order to solve the above problem, a first invention provides a character cutting means for cutting out a character pattern one character at a time from within a predetermined area containing a character string on a form, etc.; a feature extracting means for extracting features of the character pattern cut out by the character cutting means; a dictionary storage means storing a predetermined collation dictionary mask for horizontal writing, etc.;
In the character recognition device, the predetermined region is divided into an arbitrary number of regions, the character recognition device comprising: an identification device that identifies the character pattern by comparing the features extracted by the feature extraction device with the contents of the dictionary storage device. A region cutting means for dividing into a plurality of divided regions, a sub-pattern extraction means for extracting vertical sub-patterns and horizontal sub-patterns of each divided region by an arbitrary number of regions, a character direction determining means, and a converting means, It was established.

Here, the character direction determining means extracts the features of the vertical sub-pattern and the horizontal sub-pattern extracted by the sub-pattern extracting means, and determines the direction of the characters of the character string included in the predetermined area based on the features. It has a function to output the character direction determination result. Further, the converting means has a function of performing feature conversion or pattern conversion based on the result of the character direction determination, thereby causing the feature extracting means to extract the characteristics of the character pattern.

In the second to fourth inventions, the character direction determining means is configured as follows. That is, in the character direction determining means of the second invention, each time a vertical sub-pattern and a horizontal sub-pattern are extracted from the plurality of divided areas, the line width of the vertical sub-pattern and the line width of the horizontal sub-pattern are calculated, and the line width of the vertical sub-pattern is calculated. It has a function of determining the character direction by determining the sum of the line widths of the horizontal sub-pattern and the sum of the line widths of the horizontal sub-patterns, and comparing the ratio of both sums of line widths with a predetermined line width ratio reference value.

In the character direction determination means of the third invention, the line width of the vertical sub-pattern and the line width of the horizontal sub-pattern are calculated, the ratio of both line widths is determined for each divided area, and the line width ratio and the predetermined line width are determined. It has a function of determining the character direction based on the comparison result for each divided area with the line width ratio reference value.

In the character direction determination means of the fourth invention, the line width of the vertical sub-pattern and the line width of the horizontal sub-pattern are calculated, the ratio of both line widths is determined for each divided area, and the average of the line width ratio is determined in advance. It has a function to determine the character direction by comparing it with the line width ratio reference value.

A fifth invention is characterized in that in the first to fourth inventions, the predetermined area is a row.

(Operation) According to the first to fifth aspects of the invention, the character recognition device is configured as described above, so that when a form image or the like is input, the area cut-out flat step selects a predetermined area ( For example, a plurality of divided regions are generated by dividing a row) into an arbitrary number of regions, and the generation results are provided to the sub-pattern extraction means. The sub-pattern extracting means scans each of the plurality of divided regions to extract vertical sub-patterns and horizontal sub-patterns in an arbitrary number of regions, and provides the extraction results to the character direction determining means. The character direction determining means extracts the features of the vertical sub-pattern and horizontal sub-pattern extracted by the sub-pattern extracting means (for example, the line widths of the vertical sub-pattern and horizontal sub-pattern), and based on the features, performs calculation etc. to The character direction (for example, vertical direction and horizontal direction) of the character string included in is determined, and the character direction determination result is output to the conversion means.

The converting means converts the characteristics of the character pattern in the feature extracting means or converts the pattern of the character pattern based on the result of the character direction determination, thereby allowing the feature extracting means to accurately extract the characteristics of the character pattern. Then, the identification means identifies the character pattern based on the output of the feature extraction means with reference to the collation dictionary mask in the dictionary storage means. This makes it possible to accurately recognize characters in character strings on various forms, including business cards, at high speed, with good operability, and with a small memory capacity. Therefore, the above problem can be solved.

(Embodiment) FIG. 1 is a functional block diagram of a character recognition device showing an embodiment of the present invention.

Before explaining this functional block diagram, the principle of this embodiment will be explained first.

Focusing on the line width of Mincho fonts, which account for most of the type used in printed documents, we calculated the ratio of the line width of horizontal sub-patterns to the line width of vertical sub-patterns, which was obtained using the sub-pattern extraction method described later. The results of the measurements are shown in Table 1.

Table 1 Line width ratio measurement results of horizontal sub-patterns and vertical sub-patterns Figures 3(a) to (f) are diagrams showing the vertical sub-patterns and horizontal sub-patterns of the Mincho font "A". , Regarding the character "A" in the Mincho font shown in Figure 3 (a),
2 obtained by image scanning with a resolution of 37.5 μm for type size 8.12 and 16 points.
A horizontal sub-pattern and a vertical sub-pattern were extracted from the value pattern, and the horizontal sub-pattern line width and vertical sub-pattern line width were determined, respectively.

The method for extracting horizontal sub-patterns and vertical sub-patterns is to obtain the line width of the input pattern using equation (1), which will be described later, and then scan the input pattern in the horizontal and vertical directions to extract the line width from the cross-sections of the character lines detected. This is performed by extracting a cross section whose cross section is sufficiently longer than the line width of the input pattern.

An example of the vertical sub-pattern of the horizontally written "A" in the Mincho font shown in FIG. 3(a) is shown in FIG. 3(b), and an example of the horizontal sub-pattern is shown in FIG. 3(C). Also, Figure 3 (a
The vertical writing pattern examples corresponding to >, (b), and (c) are shown in the third example.
Figures (d) and (e).

(f) respectively. That is, Fig. 3 (d
) is a vertically written Mincho font "A", which is obtained by rotating the character pattern of FIG. 3(a) by 90 degrees counterclockwise. Therefore, FIG. 3(e) showing the vertical sub-pattern of FIG. 3(d) and FIG. 3(f> showing the horizontal sub-pattern) are the patterns of FIG. The pattern is the same as the one rotated 90 degrees.

As is clear from Table 1, the horizontal sub-pattern line width and the vertical sub-pattern line width of the Mincho font differ greatly. Therefore, the line width ratio (=vertical sub-pattern line width/horizontal sub-pattern line width) of horizontally written characters is 2.8 or more. Similarly, the line width ratio (=vertical sub-pattern line width/horizontal sub-pattern line width) of vertically written characters is 0.36 or less.

As can be understood from the above, it is possible to determine whether the input pattern is written horizontally or vertically depending on the line width ratio (=vertical sub-pattern line width/horizontal sub-pattern line width) of the input pattern.

Next, the configuration of the character recognition device shown in FIG. 1, which is configured based on the above principle, will be explained.

This character recognition device has an image reading means 1 that inputs characters on a form in the form of an optical signal Sin and converts it into an electrical signal by photoelectric conversion.The output side of the image reading means 1 has a random access - Image storage means 2 composed of memory (RAM) etc. is connected. The image storage means 2 stores, for example, an image for one form, and a line position detection means 3 and an area cutting means 4 are connected to its output side. The line position detection means 3 has a function of outputting the position coordinates of a line in the image storage means 2 from the form image in the image storage means 2. The area cutting means 4 has a function of dividing the line of the form image in the image storage means 2 into a plurality of areas including characters based on the output of the line position detecting means 3, and has a sub-pattern extraction function on the output side. Means 10 are connected.

The sub-pattern extracting means 10 has a function of extracting the vertical sub-patterns and horizontal sub-patterns of the plurality of divided regions cut out by the region cutting means 4 by the number of the plural regions. This sub-pattern extraction means 10 includes an area storage means 11 consisting of a buffer or the like for storing data for one area outputted from the area extraction means 4, and an area line width Wa for the area outputted from the area extraction means 4. A region line width calculating means 12 for calculating, a vertical sub-pattern extracting means 13 for vertically scanning the inside of the region storage means 11 to extract a vertical sub-pattern of the region, and a vertical sub-pattern extraction means 13 for horizontally scanning the inside of the region storage means 11 for extracting vertical sub-patterns of the region. The horizontal sub-pattern extracting means 14 extracts a horizontal sub-pattern of a region. This vertical sub-pattern extraction means 13
A character direction determining means 20 is connected to the output side of the horizontal sub-pattern extracting means 14.

The character direction determining means 20 extracts features of the vertical sub-pattern and horizontal sub-pattern, such as line width, determines whether the character direction is vertical or horizontal based on the line width, and outputs the character direction determination result. It is composed of a determining means 21 and a storage means 22 consisting of a buffer or the like.

Further, on the output side of the line position detection means 3, a character cutting means 30 is provided.
is connected. The character cutting means 30 has a function of inputting the line position coordinates of each line outputted from the line position detecting means 3 and outputting a character pattern by cutting out each character, and has a feature extraction means on the output side. 40 are connected.

The feature extraction means 40 generates a feature matrix based on the character pattern cut out by the character cutout means 30, and includes a pattern storage means 41, a line width calculation means 42, a subpattern extraction means 43, and a character frame detection means. 44, character frame division determining means 45, feature matrix extracting means 46, and feature converting means 47.

The pattern storage means 41 stores the character patterns cut out by the character cutout means 30, and provides the storage results to the sub-pattern extraction means 43 and the character frame detection means 44, and is composed of registers and the like.

The line width calculating means 42 has a function of calculating the line width W of the character pattern cut out by the character cutting means 30 and providing the calculation result to the sub pattern extraction means 43 and the feature matrix extraction means 46.

The sub-pattern extraction means 43 scans the pattern storage means 41 in a plurality of directions (for example, vertical direction, horizontal direction, right diagonal direction, left diagonal direction) and extracts vertical sub-patterns VSP, horizontal sub-patterns H3P, right-diagonal sub-patterns R3P, and left diagonal sub-patterns. It has a function of extracting the diagonal sub-pattern LSP and outputting the extraction result to the feature matrix extraction means 46. The character frame detection means 44 has a function of detecting a character frame circumscribing the character pattern in the pattern storage means 41 and outputting it to the character frame division determining means 45. The character frame division determining means 45 determines the coordinates of dividing points on the X axis (horizontal direction of the character frame) and the Y axis (vertical direction of the character frame) for dividing the internal region of the detected character frame into a plurality of regions. The determination result is output to the feature matrix extraction means 46.

The feature matrix extraction means 46 has a function of extracting the features of the character patterns in the pattern storage means 41 to generate a feature matrix, and providing it to the feature conversion means 47.

The feature converting means 47 has a function of rotating the feature matrix from the feature matrix extracting means 46 by a predetermined angle based on the character direction determination result output from the determining means 21. On the output side of this feature conversion means 47, identification means 51
is connected to the identification means 51, and a dictionary storage means 50 is connected to the identification means 51.

The dictionary storage means 50 stores, for example, a collation dictionary mask for horizontal writing, and is a read-only memory (ROM).
It is made up of etc. The identification means 51 is the feature extraction means 4
It has a function of collating the feature matrix from 7 with a collation dictionary mask and outputting the character name 5out of the collated character pattern.

Next, we will explain the operation of the character recognition device configured as above.
This will be explained with reference to FIG. 4 and FIGS. 5(a) to 5(e). In addition, Figure 4 is a diagram showing the image of a business card, Figure 5 (a
> to (e) are diagrams for explaining a binary pattern and each sub-pattern.

In FIG. 1, the optical signal Sin corresponding to the characters on the form
is input to the image reading means 1, the image reading means 1
The optical signal Sin is converted into an electrical signal and output to the image storage means 2. The image storage means 2 receives an electrical signal from the image reading means 1 and stores an image for one form. The line position detection means 3 scans the form image stored in the image storage means 2 to create horizontal and vertical peripheral distributions,
The line is cut out, and the position coordinates of the line in the image storage means 2 are output.

That is, the image of the business card as shown in FIG. 4 is stored in the image storage means 2.
If the image is stored in
The start position Ys2 and end position Ye2 of the row, and the start position Ys3 and end position Ye3 of the third row are sequentially obtained. Next, the line position detecting means 3 vertically scans the image in the image storage means 2 line by line, and for example, for the second line, the starting position XS
and the end position Xe is obtained. In this way, the line position detecting means 3 detects the line position with respect to the image in the image storage means 2, and transfers the detection result to the area cutting means 4 and the character cutting means 3.
Output to 0.

The area cutting means 4 detects the position of the line in the image storage means 2 output from the line position detecting means 3, and divides the line into a plurality of areas including characters. Taking Figure 4 as an example, there are three areas ml containing one character each of ``Kichi'', ``Den'', and ``Sei''.
, m2. Divide into m3.

That is, in this area cutting means 4, as shown in FIG.
The midpoints Xdl and Xd2 of the centers Xml, Xm2, and Xm3 of each character by vertical scanning are used as dividing points, and ml and m2 are respectively
．． Divide into three regions (divided regions) of m3. and,
The region cutting means 4 stores the data of one region in the region storage means 11 in the subpattern extraction means 10 and the region line width calculation means 1.
Output to 2.

The area storage means 11 stores 1 output from the area extraction means 4.
Data for the area is stored, and the storage results are output to the vertical sub-pattern extraction means 13 and the horizontal sub-pattern extraction means 14. The area line width calculating means 12 calculates the area line width Wa for the area output from the area cutting means 4, and outputs the line width Wa to the vertical sub-pattern extracting means 13 and the horizontal sub-pattern extracting means 14.

The area line width calculation means 12 has a shift register configuration similar to, for example, a known image processing filter circuit, and calculates the line width W using a known approximate expression as shown in the following equation.

W=1/ (1-(Q/A)) ・・・・・・
(1) In this equation (1), Q is a window (W
The number of points (this is called 4 black points) in which all points in a rectangular area are black points, and A is the number of all black points.

The area line width calculation means 12 counts the number Q of these four black dots and the number A of all black dots, and calculates the line width W from the counting results according to equation (1).

The vertical sub-pattern extraction means 13 vertically scans the area in the area storage means 11, and calculates the following based on the relationship between the length of continuous black dots and the line width Wa calculated by the area line width calculation unit 12.
The vertical sub-pattern of the area is extracted, and the extraction result is output to the determining means 21 in the character direction determining means 20. Further, the horizontal sub-pattern extraction means 14 horizontally scans the area in the area storage means 11, and determines the horizontal pattern of the area based on the relationship between the length of continuous black dots and the line width Wa calculated by the area line width calculation means 12. The sub-pattern is extracted and the extraction result is determined by the judgment means 2.
Output to 1.

The determining means 21 in the character direction determining means 20 calculates the total number of black dots and the number of four black dots for the vertical sub-pattern of the area stored in the vertical sub-pattern extraction means 13 in the same way as the area line width calculating means 12, and calculates the number of total black dots and the number of four black dots using the equation (1). The line width Wv of the vertical sub-pattern is determined using .

Furthermore, the line width wh is determined in the same manner for the horizontal sub-patterns in the area stored in the horizontal sub-pattern extracting means 14. The line width W of the vertical sub-pattern obtained in this way
v and the line width wh of the horizontal sub-pattern are stored in the storage means 22. These line width calculation processes and storage processes in the storage means 22 are performed to calculate all the areas (
For example, the three divided areas ml, m2. m3).

In the determination means 21, after the region cutting means 4 cuts out all the regions to be cut out (for example, ml, m2.m3), the vertical subpattern lines of all the regions to be cut out are stored in the storage means 22. The width Wv and the line width wh of the horizontal sub-pattern are read, and the sum of the line width Wv of the vertical sub-pattern and the line width wh of the horizontal sub-pattern is determined. Furthermore, this determining means 21 determines that the ratio ΣWv/Σwh of the total line width ΣWv of the vertical sub-patterns and the total sum Σwh of the line widths of the horizontal sub-patterns
is determined, and the ratio is compared with a predetermined line width ratio reference value wth to determine whether the following equations (2) and (3) are satisfied.

ΣWv/Σwh≧wth ・・・・・・(2)
ΣWv/Σwh<wth・curved straight 3) (in this embodiment, for example, Wth=1>) If the equation (2) is satisfied, the determining means 21 determines that the character direction is horizontal writing, and uses the horizontal writing character direction determination result as (3)
If the formula is satisfied, the character direction is determined to be vertical writing, and the vertical writing character direction determination results are output to the feature conversion means 47, respectively.

As described above, after the direction of the characters on the form image in the image storage means 2 has been determined, recognition processing is performed on each character on the form image, as described below.

First, when inputting the line position coordinates of each line outputted from the line position detecting means 3, the character cutting means 30 vertically cuts out the line at the line position in the image storage means 2 obtained by the line position detecting means 3. A peripheral distribution is created by scanning, and character patterns cut out one character at a time based on the peripheral distribution are provided to a pattern storage means 41 and a line width calculation means 42 in the feature extraction means 40. The pattern storage means 41 stores character patterns cut out by the character cutting means 30. The line width calculation means 42 calculates the number of total black dots and the number of four black dots for the character pattern output from the character cutting means 30 in the same manner as the area line width calculation means 12, and calculates the number of black dots and four black dots for the character pattern output from the character cutting means 30, and calculates the number of black dots and the number of four black dots for the character pattern outputted from the character cutting means 30, and calculates the number of black dots and the number of four black dots for the character pattern output from the character cutting means 30, The line width W of the pattern is calculated and the calculation result is sent to the sub-pattern extraction means 4.
3 and the feature matrix extraction means 46.

The sub-pattern extraction means 43 performs vertical scanning over the entire surface of the character pattern stored in the pattern storage means 41, and based on the relationship between the length of continuous black dots and the line width W calculated by the line width calculation means 42, Extract the vertical sub-pattern VSP.

Similarly, the sub-pattern extracting means 43 extracts the horizontal sub-pattern H3P by horizontal scanning, the right-diagonal sub-pattern R8P by scanning diagonally 45 degrees to the right, and the left diagonal sub-pattern LSP by scanning diagonally 45 degrees left.

An example of sub-patterns extracted in this way is shown in Figure 5 (a
) to (e). Figure 5(a) is an original pattern consisting of a binary pattern within the circumscribed character frame F, and the vertical subpattern obtained from such an original pattern is
), the horizontal sub-pattern is shown in Figure 5 (C), the right diagonal sub-pattern is shown in Figure 5 (d), and the left diagonal sub-pattern is shown in Figure 5 (e).
) are shown respectively.

A character frame detection means 44 connected to the output side of the pattern storage means 41 detects a character frame F circumscribing the character pattern in the pattern storage means 41, and outputs the detection result to the character frame division determination means 45. The character frame division determining means 45 divides the internal area of the detected character frame F into an area of MXN (however, M,
N is an integer; in this embodiment, for example, M=N=5. The coordinates of the dividing points on the X-axis and Y-axis are determined. Here, the X axis indicates the horizontal direction of the character frame F, and the Y axis indicates the vertical direction of the character frame F, respectively.

Next, the feature matrix extraction means 46 extracts vertical, horizontal, right diagonal, and left diagonal sub-patterns (VSP, H
3P, H3P and LSP) pattern storage means 41
The upper character frame area is divided into MXN areas, and the number of black dots Bij in each of the resulting divided areas is counted. This sunspot Bij
and the line width W obtained by the line width calculation means 42, calculate the feature indicating the character line length Lij by the following equation (4), and calculate M
Create an XNX four-dimensional feature matrix.

Lij=Bij/W (4) Furthermore, this feature matrix extraction means 46 extracts the horizontal subpattern by the length ΔY of the character frame F in the Y axis direction for the feature matrix of the vertical subpattern ■SP. The feature matrix of H3P is normalized by the length ΔX in the X-axis direction, and the feature matrix of right diagonal sub-pattern R3P and left diagonal sub-pattern LSP is normalized by (Δχ + ΔY)/2, and the character size is ]VIXNX4
Create a dimensional feature matrix.

The feature converter 47 converts the feature matrix extracted by the feature matrix extractor 46 based on the character direction determination result output from the determiner 21. That is, when the character direction determination result is horizontal writing, the feature matrix extracted by the feature matrix extracting means 46 is directly used by the identifying means 5.
Output to 1.

Further, when the character direction determination result is vertical writing, the feature matrix extracted by the feature matrix extraction means 46 is
Rotate clockwise and convert the horizontal feature matrix to vertical feature matrix, the vertical feature matrix to horizontal feature matrix, the left diagonal feature matrix to right diagonal feature matrix, and the right diagonal feature matrix to left diagonal feature matrix. It is output to the identification means 51.

Thereafter, the identification means 51 uses the collation dictionary mask stored in the dictionary storage means 50 to create the following equation (5) between the collation dictionary mask fj specified for each feature and the extracted feature matrix fi. ) is applied, and the category name of the dictionary mask for which this distance is the minimum value is the character name S.

Output in ut format.

This allows accurate recognition of characters on the form.

As described above, this embodiment has the following advantages.

(i) In this embodiment, the area cutting means 4 divides a predetermined area including a character string on a form into a plurality of arbitrary areas,
A subpattern extracting means 10 extracts a vertical subpattern and a horizontal subpattern from each divided area, and a character direction determining means 20 extracts a vertical subpattern and a horizontal subpattern.
extracts features of vertical sub-patterns and horizontal sub-patterns, such as line width, automatically determines the character direction of a character string within a predetermined area based on the characteristics5, and then determines the character pattern based on the character direction determination result. Extracts the features of Therefore, for various forms including business cards, the operator can omit the complicated work of specifying the character direction for each form or for each character string on a form. , it is easy to operate, and can be recognized with high precision and high speed.

(ii) In this embodiment, the region cutting means 4 divides a predetermined region into a plurality of arbitrary regions, the sub-pattern extracting means 10 extracts a vertical sub-pattern and a horizontal sub-pattern for each divided region, and the character direction determining means At step 20, the character direction is determined, and based on the determination result, feature extraction means 40 extracts the characteristics of the character pattern. Therefore, the memory (area storage means 11 and storage means 2
2 etc., it is not necessary to store the entire predetermined area, and the memory capacity of the divided area is sufficient. Furthermore, since the feature extraction means 40 extracts the characteristics of the character pattern based on the character direction determination result of the character direction determination means 20,
The number of collation dictionary masks that are rotated by a predetermined angle and stored in the dictionary storage means 50 can be reduced compared to the conventional one, thereby making it possible to reduce the memory capacity of the dictionary storage means 50. In this way, the memory capacity can be made smaller than that of the conventional device, so the entire device can be made smaller.

(iii) The character direction determining means 20 calculates the ratio between the line width Wv of the vertical sub-pattern and the line width wh of the horizontal sub-pattern, compares the ratio with a preset line width ratio reference value Wth, and determines the comparison result. Since the direction of the characters in the character string within a line is determined based on this, the configuration is simple and highly accurate character direction determination results can be obtained.

Note that the present invention is not limited to the illustrated embodiment, and various modifications are possible. Examples of such modifications include the following.

(i) In the above embodiment, the line width Wv of the vertical sub-pattern and the line width wh of the horizontal sub-pattern of each region to be cut out are
Although each region is stored in the storage means 22 and the character direction is determined after the region cutting means 4 has cut out all the regions to be cut out, it is also possible to judge the character direction by another method. It's okay.

For example, after determining the line width Wv of the vertical sub-pattern and the line width wh of the horizontal sub-pattern for each area, the ratio of the line width Wv of the vertical sub-pattern to the line width wh of the horizontal sub-pattern is determined, and the line width ratio is stored in the storage device. 22 for each area. and,
After all the regions to be cut out are cut out by the region cutting means 4, the determining means 21 determines each line width ratio and a predetermined line width ratio reference value with respect to the line width ratio for each region stored in the storage means 22. The number of areas At determined to be vertical writing and the number Ay of areas determined to be horizontal writing are determined, and the following equations (6) and (7) are determined.

At≦Ay ... Pero) At>Ay
(7) The determining means 21 determines the horizontal character direction determination result if the formula (6) is satisfied, and the vertical character determination result if the formula (7) is satisfied. It is also possible to output each of them and provide them to the feature conversion means 47.

Further, the determining means 21 calculates the average Wav of the line width ratio for each area stored in the storage means 22, compares the average Wav with a predetermined line width ratio reference value wth, and based on the comparison result, the character The direction may also be determined.

Even if the character direction determining means is configured in this manner, substantially the same advantages as in the above embodiment can be obtained.

(ii) In the above embodiment, characters are identified after the feature matrix is converted by the feature converter 47 based on the determination result of the character direction by the character direction determiner 20. For example, the character is identified by the pattern converter. Character cutting means 30
is provided on the output side of the character cutting means 30, and the pattern converting means rotates the character pattern cut out from the character cutting means 30 in a predetermined direction based on the character direction determination result of the character direction determining means 20,
The characteristics of the character pattern may be extracted based on the rotated character pattern. Even with this configuration,
Almost the same actions and effects as in the above embodiment can be obtained.

(iii) In the above embodiment, character features are extracted by finding a sub-pattern in the sub-pattern extracting means 43 and finding a feature matrix from this sub-pattern in the feature matrix extracting means 46, but with other configurations, It is also possible to extract features. For example, the pattern storage means 41 may be scanned in the horizontal and vertical directions, and the number of times a scanning line crosses the character means may be used as a feature of the character to be recognized. When configured in this way, the feature conversion means 47 converts the characters into vertically written characters.
Features obtained by horizontal scanning are output to the identification means 51 as vertical features of the character to be recognized, and features obtained in the vertical direction are output to the identification means 51 as horizontal features of the character to be recognized. It is necessary to change to

(iv) Line position detection means 3, area extraction means 4, sub-pattern extraction means 10, and character direction determination stage 2 of the above embodiment
0, the character extraction means 30, the feature extraction means 40, the identification means 51, etc. may be executed by program control by the computer G, etc., instead of being configured by individual circuits.

(Effects of the Invention) As described above in detail, according to the first invention, the area cutting means divides a predetermined area including a character string on a form into a plurality of arbitrary areas, and each A sub-pattern extraction means extracts a vertical sub-pattern and a horizontal sub-pattern from the region, a character direction determination means extracts the characteristics of the vertical sub-pattern and horizontal sub-pattern, determines the character direction of a character string within a predetermined region based on the characteristics, and Since character pattern features are extracted based on the determination results, the character direction can be determined not only for business cards but also for characters on other documents, etc., and accurate character recognition is possible based on the determination results. Become. Therefore, the operator can omit the complicated work of specifying the character direction for each form or for each character string on a form, and character recognition can be performed at high speed and with good operability.

Moreover, the region cutting means divides a predetermined region into a plurality of arbitrary regions, the subpattern extraction means extracts a subpattern for each divided region, and the character direction determination means determines the character direction based on the extracted subpattern, Since character pattern features are extracted based on the determination results, the memory capacity required for character direction determination does not need to be the entire predetermined area, and is sufficient as long as the memory capacity is the size of the divided area. It is. Furthermore, since character pattern features are extracted based on the character direction determination results, the number of data stored in the dictionary storage means can be reduced, thereby reducing the memory capacity of the dictionary storage means. Since the storage capacity can be reduced in this way, it is possible to downsize the entire device.

According to the second to fourth inventions, the character direction determining means
Since the character direction is determined based on the line width of the vertical sub-pattern and the line width of the horizontal sub-pattern, the configuration is simple and the character direction can be determined with high accuracy.

In the fifth invention, since the predetermined area is defined as a line, characters written horizontally or vertically can be accurately recognized.

[Brief explanation of drawings]

FIG. 1 is a functional block diagram of a character recognition device showing an embodiment of the present invention, FIGS. 2(a) to (d) are diagrams showing examples of printed documents, and FIGS. 3(a) to (f) are clear Figure 4 is a diagram showing an example of the vertical sub-pattern and horizontal sub-pattern of the morning typeface, Figure 4 is a diagram showing an example of a business card image, and Figures 5 (a) to (e) are diagrams for explaining the binary pattern and each sub-pattern. It is a diagram. 1... Image reading means, 2... Image storage means, 3... Line position detection means, 4...
Region cutting means, 10... sub-pattern extraction means,
12... Area line width calculation means, 13...
Vertical sub-pattern extraction means, 14...1.・Horizontal sub-pattern extraction means, 20...Character direction determination means, 2
1... Determination means, 22... Storage means,
30...Character extraction means, 40...Feature tank lJ rear stage, 46...Feature matrix extraction means, 47...Feature conversion means, 50. ...Dictionary storage means, 51...Identification means. Example of printed document Figure 2 (α) (b) (c) Hole tongue (d7) Vertical writing Akira WwM"lTK Gun (e) Peeling writing Akira I# * 1,000 sun 1 large tashi (f) Mincho typeface vertical subnomagun and horizontal subtsumatan A row 3 image of name and interest No. Figure 4

Claims (1)

[Claims] 1. Character extraction means for cutting out a character pattern one character at a time from within a predetermined region containing a character string, and feature extraction for extracting features of the character pattern cut out by the character extraction means. means, dictionary storage means storing a predetermined collation dictionary mask, and identification means for comparing the features extracted by the feature extraction means with the contents of the dictionary storage means to identify the character pattern. In the character recognition device, an area cutting means for dividing the predetermined area into an arbitrary number of areas to generate a plurality of divided areas, and extracting vertical sub-patterns and horizontal sub-patterns of each of the divided areas for an arbitrary number of areas. a sub-pattern extracting means, extracting features of the vertical sub-pattern and horizontal sub-pattern extracted by the sub-pattern extracting means, and based on the features,
character direction determining means for determining the direction of characters in a character string included in the predetermined area and outputting a character direction determination result; and feature extraction means for performing feature conversion or pattern conversion based on the character direction determination result. A character recognition device comprising: a conversion means for extracting features of a character pattern. 2. In the character recognition device according to claim 1, each time a vertical sub-pattern and a horizontal sub-pattern are extracted from the plurality of divided regions, the character direction determining means determines the line width of the vertical sub-pattern and the line width of the horizontal sub-pattern. It has a function to determine the character direction by calculating the sum of the line widths of the vertical sub-pattern and the sum of the line widths of the horizontal sub-patterns, and comparing the ratio of both sums of line widths with a predetermined line width ratio reference value. A character recognition device characterized by: 3. In the character recognition device according to claim 1, each time a vertical sub-pattern and a horizontal sub-pattern are extracted from the plurality of divided regions, the character direction determining means determines the line width of the vertical sub-pattern and the line width of the horizontal sub-pattern. The function calculates the ratio of both line widths for each divided area, and determines the character direction based on the comparison result for each divided area by comparing the line width ratio with a predetermined line width ratio reference value. A character recognition device comprising: 4. In the character recognition device according to claim 1, each time a vertical sub-pattern and a horizontal sub-pattern are extracted from the plurality of divided regions, the character direction determining means determines the line width of the vertical sub-pattern and the line width of the horizontal sub-pattern. , calculate the ratio of both line widths for each divided area, and compare the average of the line width ratios with a predetermined line width ratio reference value to determine the character direction. character recognition device. 5. The character recognition device according to claim 1, 2, 3 or 4, wherein the predetermined area is a line.