JPH0728948A - Character recognition device - Google Patents

Abstract

PURPOSE:To provide a character recognition device which can obtain the highly accurate and stable recognition performance even for a character pattern of low quality that has the local variation of line width. CONSTITUTION:Four types of subpatterns are extracted at every scanning direction based on the average line width of an original character pattern. Furthermore a blurred pattern is extracted. If it is decided that a restoration pattern must be generated for the remaining pattern obtained by excluding the infinitesimal segments form the blurred pattern, a blur restoration pattern is generated by a thickening processing. In the same way, the sub-pattern of the blur restoration pattern are extracted at every direction. Then an area included in a circumscribed frame of the original character pattern or a synthetic pattern of the original character pattern with the blur restoration pattern is divided into the lattice-shaped partial areas. Then the feature value is calculated in each divided in each divided area of either a subpattern or a synthetic subpattern of the subpattern with the blur restoration pattern. Thus a feature matrix is generated and collated with the feature matrix of a dictionary 137. Thus the characters are recognized.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a character recognition device capable of obtaining highly accurate and stable recognition performance even for a character pattern having a partially different line width such as a faint part.

[0002]

2. Description of the Related Art Conventionally, characteristics of input character patterns are extracted,
As a character recognition device that outputs a recognition result by collating with a dictionary prepared in advance, for example, Japanese Patent Publication No. 60-3875.
6 was disclosed. The outline of the processing by the character recognition device will be described below.

First, the brightness of each cell of an input character pattern is converted into a binary image which is a quantized electric signal by photoelectric conversion, and the binary image is stored in a pattern register. Then, the circumscribing frame of the character pattern in the pattern register is detected, and the line width of the character pattern in the circumscribing frame is calculated. Next, the character pattern in the circumscribing frame of the pattern register is scanned horizontally, vertically, diagonally to the right, and diagonally to the left to detect continuous black pixel components having the line width as a threshold, thereby detecting the input character pattern. Extract four sub-patterns for. Further, with respect to the character pattern in the circumscribing frame of the pattern register, non-linear division into N × M partial areas in a grid pattern in the vertical direction and the horizontal direction so that the number of black pixels in each divided area is the same. To do. Next, for each of the four types of sub-patterns, the number of black pixels of the sub-pattern in the divided partial area is counted, and the number of black pixels is normalized by the size of the character pattern to thereby distribute the character lines in each direction. An N × M × 4 dimensional feature matrix reflecting the state is extracted. Then, the feature matrix is collated with a dictionary, which is a feature matrix of a plurality of standard characters prepared in advance, and the recognition result of the input character pattern is output from the collation result.

[0004]

However, the above character recognition device has the following problems. That is,
For the binary image in the circumscribed frame of the input character pattern,
Scanning is performed in each of the four directions of horizontal, vertical, diagonal to the right, and diagonal to the left, and a stroke consisting of consecutive black pixels is extracted by using twice the average line width of the character pattern as a threshold, and the distribution thereof is represented. The seed sub-pattern was extracted. But,
The conventional sub-pattern extraction method naturally does not extract the stroke component having the number of continuous black pixels smaller than twice the line width. Therefore, a character pattern that is partly faint, that is, a character pattern whose local line width value is extremely small compared to other parts, is extracted based on the sub-pattern and sub-pattern. There was a problem that it was not reflected in the feature matrix, and as a result, it contributed to the deterioration of the recognition performance.

An example of such a case is shown in FIGS. 4 and 6 (a).
Shown in. FIG. 4 shows a case where the beard portion of the letter “Q” of the alphabet, that is, the segment in the area surrounded by the wavy line 401 is faintly divided into three.
This faint whisker is normally detected as a part of the stroke by scanning in the left diagonal direction, but in this example, as a part of the sub-pattern in scanning in any direction. Not detected. Therefore, the beard portion is not reflected in the feature, and the probability of misreading a similar character without a beard, such as "O", increases. In addition, in FIG. 6A, the kanji “Cause” is used. In this case, of the elements that make up the "cause," the inner part, "large", is written smaller than usual, with respect to the outer part, "mouth," and is less than twice the average line width of the entire character. Since it has only the size, the scan using the average line width is
“Large” as shown in (b), (c), (d), and (e)
This is a big problem because the situation where the character of “” is not reflected in any sub pattern occurs.

Further, even when the character pattern is not faint or written small as described above, the value of the average line width becomes very large due to the part being crushed, and as a result, the normal line width is increased. There is a problem in that the stroke is not detected as a sub-pattern by scanning and is therefore not reflected in the feature matrix, resulting in deterioration of recognition performance.
An example of such a case is shown in FIG. FIG. 5 is an example in which the lower part of the numeral "5" has been formed into a loop, but at this time, the average line width of the entire character becomes a large value, and as a result, for example, the broken line 501 indicates. The stroke portion or the like which is normally written as described above has a size of twice the line width or less, and is eventually not extracted as a sub pattern. Therefore, the feature is extracted as a pattern having no stroke indicated by the broken line 501, and, for example, "6".
And so on, and the probability of misreading as "6" increases.

In the conventional sub-pattern extraction method described above, when the stroke width of each stroke component forming a character is distributed near the average line width, effective feature extraction can be performed while local feature extraction can be performed. When the stroke width value of a stroke has a large difference from the stroke width value of other parts, that is, an appropriate feature for a character pattern such as a part being crushed or faint By eliminating the problem that the extraction could not be performed and the recognition performance was degraded, the local pattern width was smaller than the average line width, and a blur pattern consisting only of stroke components that was not extracted as a sub pattern in normal scanning was extracted. By creating a blur restoration pattern by performing thickening processing to these stroke components up to the average line width of the input character, the sub pattern of the blur restoration pattern is extracted and the input sentence is extracted. The feature is based on the composite sub-pattern and the divided area by synthesizing the pattern with the sub-pattern, further synthesizing the original binary image and the blur restoration pattern, and dividing the circumscribing frame from the peripheral distribution based on the synthetic character pattern. An object of the present invention is to provide a character recognition device that can obtain stable and highly accurate recognition performance even for poor quality character patterns that have large variations in local line widths by performing extraction and recognition processing. And

[0008]

In order to solve the above problems, the present invention optically scans a character pattern written on a form or the like and converts it into a binary image which is a quantized electric signal. A photoelectric conversion unit, a pattern register for storing the character pattern converted into the binary image, a circumscribing frame detecting unit for detecting a circumscribing frame of the character pattern in the pattern register, and a circumscribing frame of the pattern register. A line width calculation unit for calculating a line width of a character pattern, and a character pattern in a circumscribing frame of the pattern register is scanned in each of horizontal, vertical, right diagonal, and left diagonal directions, and a black pixel on a scanning line is scanned. A sub-pattern extraction unit that detects as strokes when the number of consecutive patterns exceeds a threshold value determined based on the line width, and extracts four types of sub-patterns representing the distribution of these strokes in each direction; Pattern register From the binary image and the four sub-patterns in the contact frame, among the black pixels constituting a character pattern,
A blur pattern extracting unit that extracts a set of black pixels that do not belong to any of the four types of sub patterns as a blur pattern,
Of the independent segments constituting the faint pattern, a fine segment removing unit that removes a fine segment, a faint pattern line width calculating unit that calculates a line width of the faint pattern, and a faint pattern from which the fine segment is removed are restored. If it is determined that a pattern needs to be created,
For each segment that constitutes the faint pattern, by performing a process of thickening the line width to the average line width, a faint pattern restoring unit that creates a faint restored pattern and a horizontal direction with respect to the faint restored pattern , Vertical, diagonal to the right, diagonal to the left, and when the number of consecutive black pixels on the scanning line exceeds the threshold value determined based on the previous average line width, it is detected as a stroke and these strokes are detected. A shading restoration sub-pattern extraction unit that extracts four types of shading restoration sub-patterns that represent the distribution of black and white in each direction, a sub-pattern synthesis unit that synthesizes the sub-pattern and the shading restoration sub-pattern for each type, and the sub-pattern or synthesis. A control unit that outputs one of the sub patterns to the feature extraction unit, a binary image in the circumscribed frame of the pattern register, and the blurring correction. A pattern synthesizing unit for synthesizing a pattern and creating a synthetic character pattern, and a circumscribing frame in a horizontal and vertical direction based on a binary image in the circumscribing frame of the pattern register or a peripheral distribution of the synthetic character pattern. Circumscribing frame dividing unit that divides the divided partial region into a shape, a feature extraction unit that calculates a feature value of the divided partial region for the sub-pattern or the composite sub-pattern, and creates a feature matrix; It has a discriminating part which outputs a final recognition result by collating with the created dictionary.

[0009]

According to the present invention, four types of sub-patterns representing the distribution of the strokes of characters based on the average line width of the original character pattern are extracted for each scanning direction, and do not belong to any of the extracted sub-patterns. A faint pattern consisting of patterns is extracted. It is determined whether or not there is a need to create a restoration pattern for the remaining patterns obtained by removing minute segments from this blur pattern, and if it is determined that it is necessary to create a restoration pattern, the blur restoration pattern is determined based on the line width of the blur pattern. The sub-pattern of the blur restoration pattern is created and extracted in each direction as described above. In addition, when the area inside the circumscribed frame of the original character pattern or the composite pattern of the original character pattern and the blur restoration pattern is divided into the grid-like partial areas, and the sub pattern or the sub pattern of the blur restoration pattern is extracted. One of the sub patterns of the combined sub patterns is output to the feature extraction unit, the feature value in the divided area is calculated, and the feature matrix is created. Character recognition is performed by comparing this feature matrix with the feature matrix of the dictionary. Therefore, even if a part of the stroke components constituting the character pattern that plays an essential role in character recognition becomes faint or small, it is possible to rescue and extract it as a faint restoration sub-pattern. As a result, it is possible to obtain a highly accurate and stable recognition performance even for a low-quality character pattern having local line width variations.

[0010]

[Embodiment] Embodiment 1 of the character recognition apparatus according to the present invention will be described below.
2 will be described, but here, for example, the element “large” and the like constituting 401 of FIG. 4, 501 of FIG. 5, and “factor” of FIG. 6A are collectively referred to as a blur pattern. To do. Further, in the first embodiment, an example in which the present invention is applied to the binary image of the Chinese character “Cause” in FIG.

FIG. 1 is a block diagram showing a first embodiment of a character recognition device according to the present invention. Here, 101 is an optical signal input obtained by scanning a character pattern with a scanner, 10
2 is a photoelectric conversion unit, 103 is a pattern register, 104 is a circumscribing frame detection unit, 105 is a character pattern line width calculation unit, 106 is a horizontal scanning unit, 107 is a horizontal sub pattern 1 memory, 1
Reference numeral 08 is a vertical scanning unit, 109 is a vertical sub-pattern 1 memory, 110 is a right diagonal scanning unit, 111 is a right diagonal sub pattern 1 memory, 112 is a left diagonal scanning unit, 113 is a left diagonal sub pattern 1 memory, and 114 is a faint pattern. An extraction unit, 115 is a blur pattern memory, 116 is a blur pattern line width calculation unit, 117 is a horizontal scanning unit, 118 is a horizontal blur sub-pattern memory, 119 is a vertical scanning unit, 1
Reference numeral 20 is a vertical blurring sub-pattern memory, 121 is a diagonal right direction scanning unit, 122 is a right diagonal blurring sub pattern memory, 1
23 is a left oblique direction scanning unit, 124 is a left oblique fading sub pattern memory, 125 is a horizontal sub pattern combining unit, 126 is a horizontal sub pattern 2 memory, 127 is a vertical sub pattern combining unit, 128 is a vertical sub pattern 2 memory, and 129 is a right oblique sub pattern. A synthesizing unit, 130 is a right diagonal sub-pattern 2 memory, 131 is a left diagonal sub-pattern 2 synthesizing unit, 132 is a left diagonal sub-pattern 2 memory, 133 is an output control unit, 134 is a line width determining unit, 135 is a feature extracting unit, 136 is an identifying unit. , 13
7 is a dictionary memory, 138 is a recognition result, 139 is a minute segment removing unit, 140 is a faint pattern restoring unit, 141 is a faint restoring pattern memory, 142 is a pattern combining unit, 14
Reference numeral 3 is a composite character pattern memory, and 144 is a circumscribing frame division unit.

First, an optical signal 101 obtained by scanning a character pattern handwritten or printed on a form with a scanner.
Is converted into an electric signal in the photoelectric conversion unit 102,
Further, it is converted into a binary image composed of binary signals by quantization and stored in the pattern register 103.

The circumscribing frame detection unit 104 is used in the pattern register 1
For the binary image stored in 03, the upper and lower ends of the binary image are detected by horizontal scanning, the left and right ends of the binary image are detected by vertical scanning, and as a result, the input character pattern is circumscribed. Get a circumscribing frame that is a phrase form. Then, the coordinate value regarding the circumscribing frame is output to the line width calculation unit 105, the horizontal scanning unit 106, the vertical scanning unit 108, the right diagonal scanning unit 110, the left diagonal scanning unit 112, and the blur pattern extracting unit 114, and the character Specify the cutout area of the pattern. In the following processing, when the binary image of the pattern register 103 is used, all the binary images within the circumscribed frame are targeted.

The character pattern line width calculation unit 105 calculates the average line width in the input character pattern. here,
As an example of how to obtain the average line width, in this embodiment, when the number of black pixels of the binary image of the character pattern in the circumscribing frame of the pattern register 103 is A and the number of black pixels is Q, the input character pattern The method of calculating the average line width Wr by the following formula was used. Wr = A / (A−Q) (1) However, 4 black pixels means that all the 2 × 2 windows become black pixels when the binary image is scanned by the 2 × 2 window. The black pixel number Q is a count of such 4 black pixels.

Next, with respect to the character pattern in the circumscribed frame of the pattern register 103, the horizontal scanning unit 106 horizontally, the vertical scanning unit 108 vertically, and the right diagonal scanning unit 110 diagonally right. , The diagonally leftward scanning unit 112 scans diagonally leftward, detects strokes that are continuous black pixels by using the value based on the line width as a threshold, and reflects their distribution state. Generate a sub pattern. At this time, the condition that the continuous black pixels are the stroke components forming the sub-pattern is given by the following equation, where L is the number of continuous black pixels. L> 2 × Wr (2) Here, Wr is the average line width of the character input pattern calculated by the character pattern line width calculation unit 105. That is, in the scanning in each direction, the stroke having a length that is more than twice the line width is extracted as the stroke that constitutes the sub-pattern in that direction. The distribution state of strokes as continuous black pixels in the circumscribing frame detected as described above is as follows: horizontal sub-pattern 1, vertical sub-pattern 1, right diagonal sub-pattern 1, left diagonal sub-pattern 1, for each scanning direction. Horizontal sub pattern 1 memory 107, vertical sub pattern 1 memory 109, diagonal right sub pattern 1 memory 111, diagonal left sub pattern 1 memory 113, respectively.
Stored in.

Taking FIG. 6 as an example, the original binary image before scanning is shown in FIG. 6A, the horizontal sub-pattern 1 is shown in FIG. 6B, the vertical sub-pattern 1 is shown in FIG. The sub pattern 1 is shown in FIG. 6 (d), and the left diagonal sub pattern 1 is shown in FIG. 6 (e). As described above, it can be seen that the element "large" that constitutes the "factor" is not reflected in each sub-pattern because it is a scale of twice the average line width or less.

The blur pattern extracting unit 114 uses the binary image in the circumscribing frame of the pattern register 103 and the horizontal sub pattern 1, the vertical sub pattern 1, the right oblique sub pattern 1, and the left oblique sub pattern 1 and is not extracted as a sub pattern. The distribution state of the stroke components is extracted as a blur pattern. FIG. 3 is a block diagram showing the configuration of the blur pattern extracting unit 114, and the inside of the frame shown by the dotted line represents the interior of the blur pattern extracting unit 114, 301 is an OR circuit unit, 302 is a memory, and 303 is A NOT circuit unit, 304 is a character pattern memory, and 305 is an AND circuit unit.

Next, the blur pattern extraction unit 1 shown in FIG.
The function of each block in 14 and the flow of processing will be described. First, the sub-pattern memories 107 and 1 in each direction
Horizontal sub-patterns 1 stored in 09, 111 and 113,
The vertical sub pattern 1, the diagonal right sub pattern 1, and the diagonal left sub pattern 1 are input to the OR circuit unit 301. In the OR circuit unit 301, the black pixel of each sub-pattern is 1, and the white pixel is 0.
Then, the pixel 1 of the sub-pattern area surrounded by the circumscribed frame
For each one, the OR logical operation of the pixel values of the four sub patterns 1 is executed, and the operation result is stored in advance in the memory 302.
Are output for each of the pixels corresponding to the same phrase-shaped area as the sub-pattern area prepared in.
The pattern that is the union of the two sub patterns 1 is the memory 302.
Generated on. This pattern is a black pixel when the pixel value of at least one sub-pattern among the four sub-patterns in each pixel of the area is 1, that is, a black pixel, and the pixel values of all four sub-patterns 1 are 0, that is, a white pixel is a white pixel. Therefore, the white pixel portion of the pattern of the union of the sub patterns was originally a white pixel in the binary image of the character pattern, or
It is either a black pixel in the binary image, but it was not extracted as a sub pattern.

Next, the NOT circuit section 303 performs a NOT operation on the pattern generated on the memory 302. The NOT circuit unit 303 sequentially converts a pixel having a pixel value of 0 into a pixel value of 1 and a pixel having a pixel value of 1 into a pixel value of 0 for each of the pixels forming the pattern on the memory 302, that is, A NOT operation for converting a white pixel into a black pixel and a black pixel into a white pixel is executed, and the operation result is stored in the memory 30.
It outputs on the said pixel in 2. As described above,
On the memory 302, a pattern in which the pattern, which is the union of the sub patterns generated by the OR circuit unit 301, is inverted in black and white is generated. On the other hand, independently of the above processing, the circumscribing frame detection unit 10 in the binary image of the pattern register 103 is
Only the binary image in the circumscribing frame detected by No. 4 is transferred to the character pattern memory 304.

Next, for the pattern on the memory 302 and the character pattern in the character pattern memory 3045, the AND circuit unit 3
An AND operation is executed by 05. In the AND circuit unit 305, for each pixel in the pattern area, an AND operation is performed between the pixel value of the pattern on the memory 302 and the pixel value of the character pattern on the character pattern memory 304 corresponding to the pixel, that is, both pixels. Only when the value is 1, the pixel value 1 is output, and when at least one of them is 0, an operation of outputting the pixel value 0 is executed, and the operation result is used as a blur pattern to make a blur. Pattern memory 115
Output to. As can be understood from the above description, this blurring pattern is obtained by extracting black pixels constituting the character pattern that do not belong to any of the black pixels of the four sub patterns 1. That is, for example, as shown by 401 in FIG. 4, a faint pattern is a stroke formed by a part of the stroke being divided into several segments.
Or a stroke originally isolated as shown by 501 in FIG. 5 and which does not reach the threshold value of twice the average line width shown in equation (2).

FIG. 6 shows the processing of this blur pattern extraction unit.
When applied to the original binary image of (a), FIG.
All black pixels of the sub-patterns of (b), (c), (d), and (e) are to be removed. Therefore, as shown in FIG. 6F, the element "large" that is not extracted as a sub-pattern. A faint pattern consisting of only

As described above, the faint pattern extracted by the faint pattern extracting unit 114 of FIG. 1 is stored in the faint pattern memory 115, but if necessary, minute segments in the faint pattern are removed. It is also possible to provide a minute segment removal unit 139 for this. For example, the following can be considered as a removal rule of the minute segment by the minute segment removing unit 139. That is, the number of contour black pixels forming the contour of each segment forming the faint pattern or the total number of black pixels of each segment is a predetermined threshold, for example, β times (β> 0) the line width Wr of the input character pattern. ) It is a rule to consider it as a minute segment when it is below. Here, the segment determined to be minute is erased on the blurred pattern memory or excluded from the subsequent processing. By deleting the minute segment as described above, it is possible to prevent deterioration of the recognition performance due to the deletion.

Next, the blur pattern line width calculation unit 116 calculates the line width of the blur pattern. As the method of calculating the line width, for example, the equation (2) used in the character pattern line width calculation unit 105 is used.

Next, in the blurred pattern restoring unit 140, the stroke components constituting the blurred pattern are thickened to the average line width of the input character pattern,
Restore the faded part. This faint pattern restoration unit 14
FIG. 7 is a block diagram showing an example of the processing in 0. In FIG. 7, a portion surrounded by a dotted line indicates the blurred pattern restoration unit 140, 701 is a segment detection unit, 702 is a segment 1 memory, 703 is a segment 2 memory, 704 is a segment N memory, and 705.
Is a contour point extraction unit, 706 is a contour black pixel addition unit, 707 is a line width determination unit, and 708 is a synthesis unit.

The stroke components constituting the blur pattern (excluding minute segments) stored in the blur pattern memory 115 are first detected by the segment detector 701.
One segment or a plurality of segments 1, 2 ,. ． , N. Then, each of the segments 1, 2 ,. ． , N are the black pixels
Converted to an address string on the faint pattern memory 115,
Segment 1 memory 702, segment 2 memory 703 ,. ． , Segment N memory 704. However, N indicates the number of segments of the blur pattern (N ≧ 1).

Next, in the contour point extraction unit 705, each segment 1, 2 ,. ． , N, all the black pixels forming the contour portion are detected and converted into address strings on the blur pattern memory 115, respectively, and then the contour point tables 1, 2 ,. ． , N
Is output to the next contour black pixel adding unit 706.

In the contour black pixel adding section 706, the segments 1, 2 ,. ． , N, with reference to the contents of the contour point table and the blur pattern memory 115, black pixels adjacent to the outside of the contour point series are added. At this time, the added black pixel is also an element constituting each segment, and accordingly, the segments 1, 2 ,. ． , N memories 702, 703, 704 are updated. When all the pixels adjacent to the outer side of the contour point series are converted into black pixels, the contour black pixel adding unit 706 again uses the newly added black pixel as the contour point series to further black the outside. Pixels are added and segments 1, 2 ,. ． , N memory 7
The contents of 02, 703, and 704 are updated. This processing is, so to speak, a method of adding one skin to the outer circumference of each segment. The same process is repeated until a stop instruction is output according to the determination result of the line width determination unit 707.

The method of converting the adjacent pixels into black pixels is performed as follows, for example. That is, 3 as shown in FIG.
Prepare a mask of × 3, and square 8 with the contour point of interest as the center.
When placed on 01, 8 squares 802, 80 other than 801
3, 804, 805, 806, 807, 808, 809
Among them, if there are white pixels, all of them are converted into black pixels.

The line width determining unit 707 always calculates a line width value that increases the blur pattern including the black pixels added by the contour black pixel adding unit 706, and the line width value is the character pattern line. When the average line width of the input character pattern calculated by the width calculation unit 105 is reached, a contour black pixel addition unit 706 outputs an instruction to stop adding black pixels, and a synthesis unit 708 outputs a synthesis start instruction.

The synthesizing unit 708, which has received the instruction to start synthesizing,
Segments 1, 2 ,. ． , N memories 702, 703, 7
On the basis of the updated address information of the black pixels 04 and the contents of the blur pattern memory 115, the blur restoration pattern with the thickened line width is synthetically output on the blur restoration pattern memory 141. At this time, by increasing the line width, black pixels protruding from the circumscribing frame region are removed.

Next, the blur restoration pattern memory 141 shown in FIG.
The horizontal scanning unit 11 with respect to the blur restoration pattern inside
7. The vertical scanning unit 119, the right diagonal direction scanning unit 121, and the left diagonal direction scanning unit 123 scan horizontally, vertically, diagonally rightward, and diagonally leftward, respectively, and black pixels consecutively exceeding a predetermined threshold are strobed. -It is detected as ku.
As a result, the sub-patterns of the blurring restoration pattern, that is, the blurring restoration sub-patterns are extracted this time, and the horizontal blurring sub pattern memory 118, the vertical blurring sub pattern memory 120, the right diagonal blurring sub pattern memory 122, and the left diagonal blurring sub pattern memory 124, respectively. Stored in.
Here, the condition for being the stroke component forming the sub-pattern is given by the equation (2).

The process here will be described with reference to FIG. 6 as an example. First, the partial pattern “large” extracted as a blur pattern from the character “factor” is shown in FIG. Fig. 6 shows the fading restoration pattern restored to the value.
It is shown in (o). The blur restoration sub-patterns obtained by scanning the blur restoration pattern horizontally, vertically, diagonally to the right, and diagonally to the left are shown in FIG.
It is shown in (h), (i) and (j). As described above, the threshold in the sub pattern extraction processing is
It is the average line width of the original binary image "cause" in (a). Figure 6
At the time of scanning the original binary image in (a), the sub-pattern of the "large" character that was not extracted because the average line width value was large,
It can be seen that the extraction is appropriately performed by the process of thickening the average line width value.

Next, the horizontal sub-pattern 1, vertical sub-pattern 1, right diagonal sub-pattern 1, left diagonal sub-pattern 1 extracted by scanning the original binary image, and the horizontal blurring restoration sub-pattern, vertical extracted by scanning for the blurring restoration pattern. A process is performed to combine the blur restoration sub-pattern, the right diagonal blur restoration sub-pattern, and the left diagonal blur restoration sub-pattern. This synthesizing process is performed by the horizontal sub-pattern synthesizing unit 1 independently for each sub-pattern in each direction.
25, the vertical sub-pattern combining unit 127, the right oblique sub-pattern combining unit 129, and the left oblique sub-pattern combining unit 131. The combined patterns are the horizontal sub-pattern 2, the vertical sub-pattern 2, the right diagonal sub-pattern 2 and the left diagonal sub-pattern 2, respectively, and the horizontal sub-pattern 2 memory 126, the vertical sub-pattern 2 memory 128, the right diagonal sub-pattern 2 memory 130, and the left diagonal sub-pattern 2, respectively. It is stored in the pattern memory 132.

Here, the synthesizing of the patterns in the synthesizing section is performed by, for example, O for each pixel of the two sub patterns.
A method of performing R calculation or the like is used. That is, in the case of combining two binary patterns, at least one of the individual pixels forming each outputs the pixel value 1, that is, the pixel value 1 if it is a black pixel, and both output the pixel value 0, That is, a composite pattern is created by a method of outputting a pixel value of 0 when it is a white pixel. In the case of FIG. 6, as the sub-pattern of the original binary image, FIG.
(C), (d), and (e) are given, and the sub patterns of the blur restoration pattern are shown in FIGS. 6 (g), 6 (h), and 6 (h), respectively.
When (i) and (j) are given, the sub-pattern 2 synthesized by the synthesizing unit is as shown in FIG.
(L), (m), and (n). Compared with the sub-pattern of the original binary image, these combined sub-patterns 2 accurately reflect the small local scale portion and are restored to the average line width value. This is naturally reflected in the feature matrix calculated based on the above, and therefore, misreading due to information loss due to conventional sub-pattern extraction can be prevented.

The sub-pattern created by the above-mentioned method is converted into a further compressed feature in the feature extraction unit 135, but in the present embodiment, the output control unit 1 is used.
33 is provided so that the sub pattern to be input to the feature extraction unit 135 can be selected. This output control unit 133
Is obtained by scanning the original binary image by stopping the restoration of the blurring pattern and the scanning for the blurring restoration pattern when the fine segment removing unit 139 determines that all the segments forming the blurring pattern are all minute. The obtained sub-pattern 1 in each direction is read from each of the memories 107, 109, 111, 113 and output to the feature extraction unit 135. Further, the line width of the blur pattern calculated by the blur pattern line width calculation unit 116 is always determined by the line width determination unit 134, and when the line width is determined to be equal to or less than a predetermined threshold value, The determination result is transmitted to the output control unit 133. At this time, for example, the following equation is given as the threshold for the line width. Ws <δ × Wr (5) 0 <δ << 1 (6) where Ws is the line width of the blurred pattern, Wr is the line width of the original binary image, and the conditions of formulas (5) and (6) are satisfied. When is satisfied, it means that Ws is extremely smaller than Wr.

Now, from the line width determination unit 134, Ws is W
The output control unit 133, which has received the determination result that it is extremely smaller than r, stops scanning for the blur pattern in the same manner as described above, and sets the sub-pattern 1 in each direction obtained by scanning the original binary image for each. Memory 10
7, 109, 111, 113, and feature extraction unit 1
To 35. The above processing of the output control unit 133 is performed in view of the problems described below.

Combining the blur-restoring sub-pattern into sub-pattern 1 obtained by scanning the original binary image restores the important information that was removed, while eliminating the non-essential stroke component of that character pattern. There is a risk of adding more. Therefore, in this embodiment, in order to remove the extrinsic stroke component, first, as described above, the minute segment of the blurred pattern before being restored in the minute segment removing unit 139 is removed, and naturally. If all the segments are judged to be very small, the original 2
Only the sub-pattern 1 obtained by scanning the value image is output to the feature extraction unit 135. Further, a line width determination unit 134 is provided, and even if the line width of the fading pattern does not reach a predetermined threshold, the fading pattern is determined to be extrinsic for recognition, and in such a case, the fading pattern is determined. Only the sub-pattern 1 is output to the feature extraction unit 135 without performing restoration and scanning. By doing so, the extrinsic stroke component is removed from the sub-pattern, and it is possible to prevent erroneous reading and the like due to it.

The feature extraction unit 135 performs feature extraction based on the sub-pattern of the input original binary image or four types of synthesized sub-patterns. Before performing this feature extraction,
In the circumscribing frame dividing unit 144, N × M is arranged in the vertical and horizontal directions in advance so that the number of black pixels in each divided region is the same as that of the character pattern in the circumscribing frame of the pattern register 103 in advance. There is a step of non-linear division into a number of sub-regions. For example, the character “Soil” shown in FIG. 9A is an example in which the circumscribing frame 901 is divided into four partial regions in each of the vertical and horizontal directions, for a total of 16 partial regions.
First, when determining a horizontal dividing line, a marginal distribution histogram projected on the Y axis as shown in FIG. 9B is obtained. The marginal distribution histogram is a histogram obtained by counting the number of black pixels existing on a scanning line parallel to the X axis for each of Y = 0 to Y = Ye (the maximum value of the Y coordinate of the circumscribing frame). Yes, the horizontal axis is the Y coordinate of the scan line,
The vertical axis represents the number of black pixels.

Here, in order to determine the position of the dividing line so that the number of black pixels in each area divided by the dividing line parallel to the X axis is the same, the following processing is performed. That is, the search for the marginal distribution histogram is started from Y = 0 and Y = 1,
Y = 2 is sequentially searched, and the number of black pixels, which is each frequency, is added. And the cumulative value of the histogram is 2
When a value obtained by dividing the total number of black pixels of the value image by the number of divisions is reached, the scanning line at that time is set as a division line. Figure 9
In (b), the dividing line 908 is first found. Similarly, starting from the scanning line next to the dividing line 908, the cumulative value is newly obtained, and the scanning line 909 in which the cumulative value reaches the above-described predetermined threshold value is detected as the second dividing line.
In this example, four divisions are adopted, so the third division line 9
When 10 is detected, the horizontal division ends. In FIG. 9A, these dividing lines 908, 909, and 910.
Correspond to 902, 903, and 904, respectively. As described above, the dividing lines 902, 903, and 904 divide the binary image into four regions each having the same number of black pixels in the horizontal direction.

As can be seen from this division example, the width of the division area is narrow in the area where the black pixels are dense, that is, in the vicinity of the maximum of the peripheral distribution histogram, and in the area where the number of black pixels is not so large. Is wide. That is, the divided area is sensitively dependent on the distribution state of the black pixels of the binary image. By dividing in this way, the deviation of the various black pixel distributions of the character pattern is alleviated, and a matrix that reflects the characteristics of the character pattern is calculated much more accurately than simple division of the circumscribing frame.

By the same method, vertical dividing lines 911, 912, 913 are detected based on the peripheral distribution histogram projected on the X axis shown in FIG. 9C, and the number of black pixels is equal to each other. Is divided into four areas. However, the dividing lines 911, 912, 913 in the vertical direction are
It corresponds to 905, 906, and 907 in FIG.

As described above, the binary image of FIG. 9A is divided into 16 partial areas by the horizontal dividing lines 902, 903 and 904 and the vertical dividing lines 905, 906 and 907. Will be.

The character pattern shown as an example in FIG. 9 (a) is a normal pattern having no fading portion, so there is no portion to be extracted as a fading pattern, or
In the minute segment removing unit 139, all minute segments are removed, it is determined that there is no blur pattern, and feature extraction based on only the sub-pattern by the first scan is executed. Therefore, the blurred pattern restoration unit 140
Since it is not subjected to the restoration processing by, the dividing line obtained from the peripheral distribution of the original binary image that follows the conventional dividing method accurately reflects the distribution of all black pixels of the target binary image. It has become a thing.

However, if a conventional division method is applied to the original binary image as it is with respect to a character pattern that is partially extracted as a blurred pattern and is then restored, there arises a problem. For example, consider a character pattern as shown in FIG. 10A in which the stroke component in the vertical direction is faint in the character "soil" in FIG. 9A. Figure 10 (a)
10B and 10C are marginal distribution histograms projected on the Y axis and the X axis of the binary image of FIG.
The horizontal dividing lines obtained by the method described above are given by 1008, 1009, 1010, and the vertical dividing lines are given by 1011, 1012, 1013. However, these dividing lines are 1002, 1003, 1004, and 100 in FIG.
5, 1006, 1007. At this time, these dividing lines are set to positions slightly different from the case where the dividing lines are not blurred because the vertical stroke is blurred and the number of black pixels is reduced.

For example, in FIG. 10 (c), the width of the area partitioned by the dividing lines 1012 and 1013 is wider because it is necessary to obtain more black pixels than when the vertical stroke component is not blurred. Has become. In the conventional method, the dividing line 100 obtained for the original binary image as described above is used.
2, 1003, 1004 and 1005, 1006, 10
The circumscribing frame is divided by 07, and the features described later are obtained for each partial area. However, when the characteristics of the restored combined sub-pattern are calculated using the divided areas determined based on the original binary image, the recognition performance is poor because the correspondence between the divided areas and the sub-patterns is different as described above. The problem of lowering occurs.

In the present invention, in order to solve the above problems, the circumscribed frame is not divided into the original binary image, but
The original binary image and the blur restoration pattern are combined to create a combined character pattern, and the circumscribing frame is divided based on the peripheral distribution of the combined character pattern. As a result, there is no discrepancy between the combined sub-pattern and the divided area, and it is possible to prevent the deterioration of recognition performance. This is one of the major features of the present invention.

In order to realize the above method, in the first embodiment, the pattern synthesizing unit 142 and the synthetic character pattern memory 14 shown in FIG.
3 is provided. First, in the pattern synthesizing unit 142, the original binary image in the circumscribing frame of the pattern register 103 and the blur restoration pattern in the blur restoration pattern memory 141 are synthesized, and the synthesized character pattern memory 14 is used as a synthesized character pattern.
Output to 3. And for this composite character pattern,
The circumscribing frame dividing unit 144 performs the dividing process to obtain a divided area. Here, the composition in the pattern composition unit 142 is the same as the processing performed in the sub-pattern composition unit or the like.

For example, let us consider an example in which the present invention is applied to the character "earth" in which the stroke component in the vertical direction is faint in FIG. When this faint stroke component is restored, the strokes become thicker and connect, so that the composite character pattern becomes a pattern similar to that in FIG. 9 (a). Here, suppose that FIG.
Assuming the composite character pattern of (a), the circumscribing frame dividing unit 144
The dividing line obtained by
3, 1004, 1005, 1006, 1007 were replaced with 902, 903, 904, 905, 906, 9
07, and the coordinate values of these dividing lines are the feature extraction unit 13
5 is output. It should be noted that when the fading pattern restoration processing and the sub-pattern extraction processing are not performed, naturally, the synthetic character pattern is not created, and the pattern register 1
The dividing line for the binary image in the circumscribed frame of 03 is detected,
Those coordinate values are output to the feature extraction unit 135.

Next, for each of the four types of sub-patterns 1 or the combined sub-pattern 2, the number of black pixels of the sub-pattern in the divided partial area is counted,
By normalizing this with the size of the character pattern,
N × M × 4 reflecting the distribution of character lines in each direction
The dimensional feature matrix is extracted and output to the identification unit 136.

The identifying unit 136 collates the feature matrix with a feature matrix of a plurality of standard characters stored in advance in the dictionary memory 137, and finally, the candidate category narrowed down to one is selected as the input character pattern. Is output as the recognition result 138.

Example 1 was an extremely effective method when the local line widths of strokes forming characters and figures could be classified into two. However, it can be considered that ordinary simple characters can be almost reflected in the sub-pattern even when scanning with two kinds of line widths, while there are two steps for complicated figures and kanji, etc. consisting of strokes of three or more kinds of scales. There may be stroke components that cannot be captured by scanning. The second embodiment has been invented in view of such a problem, and the first embodiment is a scanning with a two-step line width, whereas the second embodiment is a scan.
Is further generalized to enable M stages (M ≧ 2) of scanning. The second embodiment will be described below.

FIG. 2 is a block diagram showing a second embodiment according to the present invention. Here, 201 is an optical signal input, 202 is a photoelectric conversion unit, 203 is a pattern register, 204 is a circumscribing frame detection unit, 205 is a register, 206 is a line width calculation unit, and 207.
Is a horizontal scanning unit, 208 is a horizontal pattern memory, 209
Is a vertical scanning unit, 210 is a vertical pattern scanning unit, 211
Is a right diagonal scanning unit, 212 is a right diagonal pattern memory, 2
13 is a left oblique direction scanning unit, 214 is a left oblique pattern memory, 215 is a blurred pattern extraction unit, 216 is a minute segment removal unit, 217 is a line width determination unit, 218 is a horizontal pattern composition unit, 219 is a horizontal composition pattern memory, 220 is a vertical pattern composition unit, 221 is a vertical pattern pattern memory, 222
Is a right diagonal pattern synthesis unit, 223 is a right diagonal synthesis pattern memory, 224 is a left diagonal pattern synthesis unit, 225 is a left diagonal synthesis pattern memory, 226 is a loop counter, 227 is an output control unit, 228 is a feature extraction unit, 229 Is an identification unit, 230
Is a dictionary memory, 231 is a recognition result, 232 is a blurred pattern restoring unit, 233 is a pattern combining unit, 234 is a combined character pattern memory, and 235 is a circumscribing frame dividing unit.

Here, differences from the first embodiment will be mainly described. First, 201, 202, 203, 20
In No. 4, according to the first embodiment, only the data within the circumscribing frame of the binary image of the pattern register 203 is transferred to the register 205. As will be described later, not only the binary data of the character pattern but also the blurred pattern are sequentially overwritten in the register 205. The line width calculation unit 206 uses this register 2
The line width is calculated for the data in 05. Since the binary data of the character pattern is currently stored, the average line width of the character pattern is calculated. The method of Example 1 is also applied to the calculation of the line width.

Next, as in the first embodiment, this register 2
For the binary data in 05, the horizontal scanning section 207,
The vertical scanning unit 209, the right diagonal scanning unit 211, and the left diagonal scanning unit 213 scan horizontally, vertically, diagonally to the right, and diagonally to the left, respectively, and the subpatterns are extracted with the line width as a threshold. The data is stored in the horizontal pattern memory 208, the vertical pattern memory 210, the right diagonal pattern memory 212, and the left diagonal pattern memory 214.

Next, the blur pattern extracting unit 215 extracts the blur pattern from the binary data of the character pattern of the register 205 and the sub patterns stored in the memories 208, 210, 212 and 214 and transfers it to the register 205. To do. At this time, the extraction of the faint pattern is performed as shown in FIG.
The fading pattern extraction processing shown in FIG. 2 is performed, and this fading pattern is referred to as a fading pattern 1 for convenience. Then, the minute segment removing unit 216 removes the minute segment of the blurred pattern 1, and after checking the number of remaining segments, the line width calculating unit 206 calculates the line width of the blurred pattern 1 and further determines the line width. The unit 217 determines whether to scan the fading pattern 1 based on the line width value. However, the determination by the minute segment removal unit 216 or the line width determination unit 217 is applied to the first embodiment.

If it is determined that the blur pattern 1 does not need to be scanned, the memories 208, 210,
The sub patterns stored in 212 and 214 are output to the feature extraction unit 228 through the output control unit 227, and when it is determined that scanning is necessary, the horizontal composition pattern memory 219 and the vertical composition pattern memory 221, respectively. Right diagonal composition pattern memory 223, left diagonal composition pattern memory 2
25. In FIG. 3, the memory 107,
109, 111, and 113 are the memory 20 in FIG.
8, 210, 212, and 214, and the character pattern memory 304 is replaced with the register 205.

Next, in the same manner as in the first embodiment, for the blur pattern 1 in the register 205, the blur pattern restoring unit 232 is executed.
At, a blur restoration pattern 1 is generated. The blur restoration pattern 1 is used for the scanning units 207, 20 in each direction.
9, 211, and 213 are again scanned, and the sub-pattern of the blurring restoration pattern 1, that is, the blurring restoration sub-pattern 1, is extracted based on the average line width, and the memory 208, respectively.
210, 212, and 214 are stored.

Next, in the horizontal pattern synthesizing unit 218, the vertical pattern synthesizing unit 220, the right diagonal pattern synthesizing unit 222, and the left diagonal pattern synthesizing unit 224, the memories 208, 210, 21.
The blur restoration sub-pattern 1 stored in 2, 214 and the sub-patterns stored in the memories 219, 221, 223, 225 are combined, and the combined sub-pattern 1 is output to the memories 219, 221, 223, 225 again. The synthetic sub-pattern 1 is the same as that synthesized as a result of extracting the sub-patterns twice in the first embodiment.
However, in the second embodiment, the character pattern in the circumscribing frame of the pattern register 203 is transferred to the register 205 again, and the blur pattern extracting unit 215 uses the binary data of this character pattern and the sub pattern 1 synthesized, It is possible to extract the stroke component that has not been detected by the second scan and store it in the register 205 as the blur pattern 2. Here, the memory 1 in FIG.
07, 109, 111 and 113 correspond to the synthetic pattern memories 219, 221, 223 and 225 in FIG.
The character pattern memory 304 is replaced with the register 205.

Next, with respect to the blur pattern 2, the blur pattern restoring unit 232 performs restoration processing up to the average line width to create a blur restoration pattern, and the blur restoration sub-pattern 2 is set by scanning with the average line width as a threshold. The synthesizing unit 218,
In 220, 222, 224, the memories 219, 22
1, 223, 225 are combined with the combined sub-pattern 1, and the memories 219, 221, 223, 22 are again combined.
It outputs to 5 as synthetic sub-pattern 2. In exactly the same manner, the blur restoration pattern K is created for the blur pattern K, the blur restoration sub-pattern K is obtained by scanning with the average line width as a threshold, and the combining units 218, 220, 222, and 2 are performed.
24, memories 219, 221, 223, 225
The synthesized sub-pattern K-1 stored in the memory is stored in the memory 219, and the synthesized sub-pattern K-1 is output again to the memories 219, 221, 223 and 225.

The loop counter 226 counts the number of times K of sub-pattern combination, and when K reaches a predetermined threshold M, notifies the output control section 227 of this. At that time,
In the output control unit 227, the memories 219, 221, 22
The combined sub-pattern M stored in Nos. 3 and 225 is transferred to the feature extraction unit 228. Even when the number of times of composition K does not reach M, if the minute segment removal unit 216 or the line width determination unit 217 determines that there is no need to restore and scan the fading pattern K, the composition sub-pattern K at that point in time. Is transferred to the feature extraction unit 228.

Further, in the pattern synthesizing unit 233, the blur restoration patterns 1, 2, ..., Which are output to the register 205 based on the binary image in the circumscribing frame of the pattern register 203. ． , K
Are sequentially combined and output to the memory 234 as a combined character pattern. When the output control unit 227 outputs the combined sub-pattern M to the feature extracting unit 228, the circumscribing frame dividing unit 235 causes the circumscribing frame dividing unit 235 to store the original binary image and the blur restoration patterns 1, 2 ,. ． , M, the circumscribing frame division is performed based on the peripheral distribution of the combined character pattern, and the division coordinates are notified to the feature extraction unit 228.

The feature extraction unit 228, the identification unit 229, the dictionary memory 230, and the recognition result 231 are all the same as those in the first embodiment, and therefore their explanations are omitted.

As described above, according to the second embodiment, by scanning M times, it is possible to create a sub-pattern that reflects M kinds of stroke components having different line widths. Highly accurate recognition performance can be stably maintained even for complicated Chinese characters and figures. Further, the first embodiment is equivalent to the case where M = 1 in the second embodiment, and corresponds to the special case of the second embodiment.

The first and second embodiments are not limited to the above examples. For example, the blur pattern extracting means in the blur pattern extracting unit 114 or 215 is not limited to the method shown in FIG.
There are several possible methods of outputting the same result by combining ND, NAND, NOT circuits, etc., but if any of the methods can extract the blur pattern defined in the present embodiment, all of them can be used in the present invention. Belong to.

In FIG. 7, as a process for thickening the stroke of the fading pattern to the average line width, a means for sequentially adding black pixels to the outside of the contour point series has been taken. If so, there is no particular need to limit to this method, and it can be set arbitrarily.

Further, as a method of removing the extrinsic stroke component, the removal of the minute segment, the judgment by the line width, etc. were used, but the conditional expressions and the setting of the threshold value, etc. are within the scope of the present invention. It can be changed arbitrarily.

The line width calculation method, the feature matrix extraction method, the circumscribing frame division method, etc. can be appropriately changed within the scope of the present invention. Further, in the block diagrams of FIG. 1, FIG. 2, and FIG. 7, the configuration of the pattern register and each memory, the processing and operation shared by each component, the flow of input / output signals, the number of installed units,
The position and other conditions can be arbitrarily changed.

[0068]

As described above in detail, according to the present invention, the input character pattern is converted into a quantized binary image, the circumscribing frame of the binary image is detected, and the circumscribing frame is detected. The line width of the value image is calculated, and the binary image in the circumscribing frame is scanned horizontally, vertically, diagonally to the right, and diagonally to the left to determine the distribution state of strokes exceeding twice the line width. Four types of sub-patterns to be reflected are extracted, the binary image in the circumscribing frame and the four types of sub-patterns are used to detect the fading patterns not extracted as sub-patterns, the line width of the fading patterns is calculated, and the fading patterns are calculated. The line width of the blur pattern is increased to the average line width of the input character pattern by sequentially adding black pixels to the outside of one or a plurality of segments that do not contact each other to form a blur restoration pattern. Create and do more 4 types that reflect the detected stroke distribution state by scanning horizontally, vertically, diagonally to the right, and diagonally to the left by using a threshold value set based on the average line width for the restoration pattern. The subtle restoration sub-pattern of is extracted, and the sub-pattern and the sub-restoration sub-pattern are synthesized for each type to create a synthetic sub-pattern, and further, the binary image in the circumscribing frame and the blur restoration sub-pattern are synthesized. As a composite character pattern, the circumscribing frame is divided based on the composite character pattern, the feature matrix is extracted based on the division information of the synthesizing sub-pattern and the circumscribing frame, and as a result of collating the feature matrix with the dictionary, Since the recognition result is output, a part of the stroke that constitutes the character pattern and plays an essential part in the recognition is different from the other parts. Even if it becomes smaller than the local line width, it is extracted as a part of the sub-pattern and is restored until it has the average line width, and the restoration effect is reflected in the divided areas as well. It is possible to prevent deterioration of recognition performance due to loss. Therefore, it is possible to stably maintain high-precision recognition performance even for complicated Kanji characters and figures composed of poor quality character patterns with large differences in local line widths and strokes of various scales. A character recognition device can be realized.

[Brief description of drawings]

FIG. 1 is a block diagram showing a first embodiment of a character recognition device.

FIG. 2 is a block diagram showing a second embodiment of the character recognition device.

FIG. 3 is a block diagram showing a configuration of a blur pattern extracting unit.

FIG. 4 is a diagram showing an example of a pattern in which a blurred portion exists.

FIG. 5 is a diagram showing an example of a pattern that has a portion that is not extracted as a sub-pattern due to crushing.

FIG. 6 is a diagram showing an application example of the present invention.

FIG. 7 is a block diagram showing a configuration example of a blur pattern restoration unit.

FIG. 8 is a diagram showing a 3 × 3 mask.

FIG. 9 is a diagram showing an example of character frame division.

FIG. 10 is a diagram showing an example of character frame division.

[Explanation of symbols]

 101 optical signal 102 photoelectric conversion unit 103 pattern register 104 circumscribing frame detection unit 105 character pattern line width calculation unit 106 horizontal direction scanning unit 107 horizontal sub pattern 1 memory 108 vertical direction scanning unit 109 vertical sub pattern 1 memory 110 right diagonal direction scanning unit 111 right Diagonal sub-pattern 1 memory 112 Left diagonal sub-pattern scanning unit 113 Left diagonal sub-pattern 1 memory 114 Blurred pattern extraction unit 115 Blurred pattern memory 116 Blurred pattern line width calculation unit 117 Horizontal scanning unit 118 Horizontal blur sub-pattern memory 119 Vertical scanning unit 120 Vertical Faint sub-pattern memory 121 Right diagonal scan section 122 Right diagonal sub-pattern memory 123 Left diagonal scan section 124 Left diagonal sub-pattern memory 125 Horizontal sub-pattern combining section 126 Horizontal sub-pattern 2 Memory 127 Vertical sub-pattern combining unit 128 Vertical sub-pattern 2 memory 129 Right oblique sub-pattern combining unit 130 Right oblique sub-pattern 2 memory 131 Left oblique sub-pattern combining unit 132 Left oblique sub-pattern 2 memory 133 Output control unit 134 Line width determination unit 135 Feature extraction unit 136 Identification Part 137 Dictionary memory 138 Recognition result 139 Small segment removal part 140 Blurred pattern restoration part 141 Blurred restoration pattern memory 142 Pattern composition part 143 Composite character pattern memory 144 Encircling frame division part

Claims (1)

[Claims] 1. A photoelectric conversion unit for optically scanning a character pattern written on a form or the like to convert it into a binary image which is a quantized electric signal, and a character pattern converted into the binary image. , A circumscribing frame detecting unit that detects a circumscribing frame of the character pattern in the pattern register, a line width calculating unit that calculates a line width of the character pattern in the circumscribing frame of the pattern register, and the pattern. For the character pattern in the circumscribed frame of the register,
Scanning is performed in each of horizontal, vertical, right diagonal, and diagonal left directions, and when the number of consecutive black pixels on the scanning line exceeds a threshold value determined based on the line width, it is detected as a stroke, and these strokes are detected. -A sub-pattern extraction unit that extracts four types of sub-patterns representing the distribution of black and white in each direction, and a binary image in the circumscribed frame of the pattern register and the four types of sub-patterns among the black pixels that form a character pattern. A blur pattern extraction unit that extracts a set of black pixels that do not belong to any of the four types of sub patterns as a blur pattern, and a fine segment removal unit that removes a fine segment from each of the independent segments that form the blur pattern. , A blur pattern line width calculation unit that calculates the line width of the blur pattern, and a restoration pattern needs to be created for the blur pattern from which minute segments have been removed. When it is determined, for each segment that constitutes the faint pattern, a faint pattern restoring unit that creates a faint restored pattern by performing a process of thickening the line width to the average line width, and the faint restore Horizontal, vertical, diagonal to the right,
Scanning is performed in each diagonal left direction, and when the number of consecutive black pixels on the scanning line exceeds a threshold value determined based on the previous average line width, it is detected as a stroke, and the distribution of these strokes is represented. A shading restoration sub-pattern extraction unit that extracts four types of shading restoration sub-patterns for each direction, a sub-pattern synthesis unit that synthesizes the sub-pattern and the shading restoration sub-pattern for each type, and one of the sub-patterns or synthesis sub-patterns. A control unit that outputs a sub-pattern to a feature extraction unit, a pattern synthesizing unit that synthesizes a binary image in the circumscribing frame of the pattern register and the blurring restoration pattern, and a composite character pattern; and a circumscribing frame of the pattern register. The circumscribing frame is divided into grid-like partial regions in the horizontal and vertical directions based on the binary image in the image or the peripheral distribution of the composite character pattern. A circumscribing frame dividing unit, a feature extracting unit that calculates a feature value of the divided partial region for the sub-pattern or the composite sub-pattern, and creates a feature matrix, and collates the feature matrix with a dictionary prepared in advance. A character recognition device having a discriminating section for outputting a final recognition result.