JPH06274690A - Character recognizing device and optical character reader - Google Patents

Abstract

PURPOSE:To quickly correct image data and to recognize a character with high accuracy by dividing an area formed with an internal boundary frame and an external boundary frame into plural areas, and performing correction processing selectively after detecting the area for which the correction processing is required. CONSTITUTION:A correction part 16 is constituted of an internal boundary correction part and an external boundary correction part. The image data in field memory 17 is corrected by applying internal boundary frame correction processing to an internal small area designated by a control part 15 in page memory 11 at the internal boundary frame correction part similarly as a conventional device. Similarly, at the external boundary frame correction part, the image data in the memory 17 is corrected by applying external boundary frame correction processing to an external area designated by the control part 15 in the memory 11. A described character can be reproduced with high accuracy and the correction processing can be accelerated by correcting the image data in the memory 17 in such way.

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 for recognizing image data (image data, the same applies hereinafter) such as characters and numbers as character data, and image data read by optically scanning a form or the like. The present invention relates to an optical character reading device that performs character recognition based on.

[0002]

2. Description of the Related Art Image data is generated by optically scanning characters or numbers written on a form or the like, or arbitrarily set symbols (hereinafter simply referred to as "characters"), and the image data is converted into characters. There is an optical character reader (OCR) as a device that recognizes with high accuracy.

FIG. 9 is a block diagram of a conventional character recognition device included in this type of optical character reading device. In FIG. 9, 91 is a page memory that stores image data to be recognized, 92 is a frame detection / removal unit, 93 is a field memory, 94 is a character cutout unit, and 95 is a character recognition unit.

FIG. 10 shows an example of a form image (a part of a field provided with a character frame) which is a recognition object. As shown in FIG. 10, a recognition target such as a form is provided in advance with a character entry frame (a non-dropout color, simply described as a frame) having a color similar to that of the character.

The page memory 91 stores form image data obtained by optically scanning an object to be recognized. An XY coordinate system is set on the page memory 91. For example, the X-axis direction is virtually set as the row direction and the Y-axis direction is set as the column direction.

The frame detection / removal unit 92 obtains a projection histogram (black pixel cumulative number) in the row and column directions for the form image on the page memory 91 as shown in FIG.
The frame position is detected based on that. Next, an internal boundary frame is set inside the frame position, and the frame area formed by this is cut out as an area containing image data.

The field memory 93 stores the image cut out by the frame detection / removal unit 92. An XY coordinate system is set on this memory, and for example, the X-axis direction is virtually set as the row direction and the Y-axis direction is set as the column direction.

The character cut-out unit 94 takes projections in the row and column directions with respect to the image on the field memory 93, and cuts out and outputs character patterns based on these projections.

The character recognition unit 95 is provided with the character cutout unit 94.
The character is identified based on the character pattern from and the recognition result is output.

FIG. 12 shows an explanatory diagram of a method of detecting a frame and setting an internal boundary frame by the frame detecting / removing unit. Normally, this frame has a rectangular shape including an upper side, a right side, a left side, and a lower side. First, as shown in FIG. 11, the X-axis direction is the row direction,
A projection histogram in the row direction when the Y axis direction is the column direction is obtained. When this projection histogram is viewed in the column direction, the case where the value first becomes equal to or more than the threshold THL is set as the upper end of the frame upper side, and the position at which the value becomes equal to or less than the threshold THL is set as the lower end of the frame upper side. Next, at the bottom of the detected frame, a predetermined margin'WL '
The position to which is added is the upper side of the internal boundary frame. Similarly, the right side, the left side, and the lower side of the frame are designated to determine the internal boundary frame.

FIG. 13 shows an explanatory view of a character string image cut out by the frame detection / removal unit. As shown in this figure, if the frame area is simply cut out, a portion where the character and the frame are in contact with each other or a portion where the character protrudes from the frame are missing. To make up for such a missing part,
As shown in FIGS. 14 and 15, after the frame area is simply cut out, the inner boundary frame and the outer boundary frame are corrected.

FIGS. 14A, 14B, and 14C are explanatory views of the internal boundary frame correction processing.

First, as shown in FIG. 14A, the form image in the page memory 91 is searched for the presence or absence of image data in the frame area. Normally, image data to be a character recognition target is composed of black pixels, and therefore the presence or absence of black pixels is detected to search the image data. Next, FIG.
As shown in FIG. 4 (b), black pixel positions A1 and A2 in the frame area in the page memory 91 are detected, points corresponding to these positions A1 and A2 in the field memory 93 are detected, and the spaces between them are complemented to form black dots. Turn into. As a result, the image data as shown in FIG. 14C is stored in the field memory 93.

15 (a) to 15 (e) are explanatory views of the external boundary frame correction processing. As shown in FIG. 15A, the form image in the page memory 91 is searched for a black pixel in each area formed by the outer boundary frame and the inner boundary frame.
The outer boundary frame is set outside the frame, and the position where the predetermined margin'WL 'is added to the upper end of the frame is the upper position of the outer boundary frame, and the right position is the same as the inner boundary frame. ,
It is determined by specifying the left side position, the lower side position, and the like. Next, as shown in FIG. 15B, the page memory 9
1, the black pixel position B1 in the area formed by the outer boundary frame and the black pixel position B2 formed by the inner boundary frame are detected, and these positions B1, B in the field memory 93 are detected.
The points corresponding to 2 are detected, and the spaces between them are complemented to form black dots (FIG. 15C). Further, as shown in FIG. 15D, the position B1 on the external boundary in the page memory 91 is
The portion protruding from the outer boundary frame is traced with the start position as the start position, the point corresponding to this position B1 is detected in the field memory 93, and the inside of the trace range is complemented (FIG. 15 (e)).

As described above, after the image data on the field memory 93 is corrected, the image data is cut out by the character cutting section 94, and the character recognition section 95 performs character recognition to improve the character recognition accuracy. I am trying.

[0016]

However, in the above-mentioned optical character reading device, although the processing time is shortened if the frame area is simply cut out, the characters are in contact with the frame or are out of the frame. If it is filled in, the cut-out character pattern may overlap the frame, or part of the protruding character may be lost. Therefore, there is a drawback that the character recognition accuracy decreases.

Further, when the correction is performed in consideration of the frame contact and the frame protrusion as described above, it is formed by the outer boundary frame and the inner boundary frame regardless of whether or not the contact between the character and the frame occurs. The detailed removal process is performed on the entire region to be removed. Therefore, even if there is little contact or protrusion of characters, there is a problem that the start timing of the character recognition process is delayed, and as a result, the speed of the entire reading process decreases.

The present invention has been made under the background described above, and an object thereof is to provide a character recognition device for recognizing image data to be recognized as character data at high speed and with high accuracy.

Another object of the present invention is to provide an optical character reader which enables high-speed and highly accurate character recognition by using this character recognition device and an optical image input device.

[0020]

A character recognition device of the present invention which achieves the above object, is a character entry frame detecting means for detecting a character entry frame in image data to be processed, and a character entry frame detected. Boundary frame setting means for setting an inner boundary frame and an outer boundary frame respectively inside and outside, and a frame area formed by the inner boundary frame and the outer boundary frame and the character entry frame is divided into a plurality of small areas, respectively. Flag data generating means for generating flag data indicating the presence or absence of image data in each small area, and correction means for determining a correction area based on the flag data and correcting image data existing in the correction area. And a character data recognition unit that recognizes the corrected image data as character data, and determines the correction area to selectively remove the frame.

In this structure, the flag data generating means includes a first means for detecting the presence or absence of image data in each internal small area formed by dividing the internal boundary frame for each side, and the external means. From a second means for detecting the presence or absence of image data in each external small area formed by dividing the boundary frame for each side, and a third means for determining the correction area according to each detection result. Become.

Further, the optical character reader of the present invention which achieves the above-mentioned other objects, is an optical image for optically scanning a recognition object to generate image data including a character entry frame in which characters are entered. In an optical character reader having an input device and a character recognition device for recognizing the character based on the image data, the character recognition device of the present invention is used as the character recognition device.

[0023]

In the character recognition device of the present invention and the optical character reading device using the same, the frame area formed by the inner boundary frame and the outer boundary frame is divided into a plurality of small areas, and the small area is divided into a plurality of small areas. Detects whether correction processing is necessary. Based on the result, a small area to be corrected is determined and the frame is selectively removed. Therefore, the image data can be corrected promptly as compared with the case where detailed correction such as frame removal is performed on the entire area, and the correction accuracy does not decrease. In particular, when the area requiring the correction processing is not detected, the character recognition processing is immediately performed without performing the correction processing, so that the processing time is significantly shortened.

[0024]

EXAMPLE In this example, character recognition of image data obtained by optically scanning a form was performed by a character recognition device. A block configuration diagram of this character recognition device is shown in FIG. In FIG. 1, 11 is a page memory, 12 is a frame detection unit (character entry frame detection means), 13 is a boundary frame setting unit (boundary frame setting means), 14 is a flag data generation unit (flag data generation means), and 15 is control. Reference numeral 16 is a correction unit (correction unit), 17 is a field memory, 18 is a character cutout unit, and 19 is a character recognition unit (character data recognition unit).

The page memory 11, the field memory 17, the character cutout unit 18, and the character recognition unit 19 have the same functions as the corresponding parts 91, 93, 94, and 95 of the above-mentioned conventional apparatus.

The frame detector 12 detects a frame from the image data of the form stored in the page memory 11 in the same manner as the conventional device. The boundary frame setting unit 13 sets the inner boundary frame and the outer boundary frame as in the conventional device. In the present embodiment, a rectangular frame is used, and similarly to the conventional apparatus, a rectangular inner boundary frame and outer boundary frame having an upper side, a right side, a left side, and a lower side are set. Further, the image data in the frame area formed by the internal boundary frame is output to the field memory 17. The method for detecting the frame is not particularly limited, and detection can be performed by any method. Also,
The shapes of the inner boundary frame, the outer boundary frame, and the frame are not particularly limited, and any shape can be used.

As shown in FIG. 2, the flag data generating unit 14 is composed of an inner boundary frame dividing unit 141, an outer boundary frame dividing unit 142, and an image data detecting unit 143.

As shown in FIG. 3, the internal boundary frame dividing section 141 detects four vertices of the rectangular internal boundary frame, that is, an upper left vertex, an upper right vertex, a lower right vertex and a lower left vertex. Then, the frame is divided into four parts of an upper side, a lower side, a left side, and a right side according to the vertical and horizontal directions of these respective vertices. At the upper left apex of the inner boundary frame, the upper side and the left side exist respectively to the right and left of this point. Therefore, the upper side of the inner boundary frame is designated as the right direction and the left side of the inner boundary frame is designated as the lower direction. Similarly,
Also at other vertices, the upper, lower, left and right sides of the inner boundary frame are specified by the upper, lower, left and right directions. FIG. 4 shows the relationship between the specified directions at the vertices of the internal boundary frame. In this figure, it is shown that the side in the direction marked with a circle is designated at each vertex. As a result, an internal small area including each side is specified.

The outer boundary frame dividing unit 142 also detects the top, bottom, left and right vertices of the outer boundary frame as shown in FIG. 5 in the same manner as the inner boundary frame dividing unit 141, and at each vertex is shown in FIG. Specifies the upper, lower, left, and right sides of the outer bounding box in the specified direction. As a result, the external small area including each side is specified.

The image data detector 143 detects the presence or absence of image data on all sides of the inner boundary frame designated by the inner boundary frame divider 141, and the contact presence / absence flag is detected based on the result. To generate. The contact presence / absence flag indicates “present” if image data is detected, and “absent” if there is no image data.

Similarly, the image data detection unit 143 detects the presence or absence of image data on all sides of the outer boundary frame, and generates an overflow flag based on the result.

The protrusion presence / absence flag indicates "present" when image data is detected, and "absent" when there is no image data (above, first and second means).

In this embodiment, the image data is stored as a black pixel and the presence or absence of this black pixel is checked.
The presence or absence of image data can be checked by any method depending on the conditions.

As shown in FIG. 7, the correction unit 16 is composed of an inner boundary frame correction unit 161 and an outer boundary frame correction unit 162.

The control unit 15 selects all the small areas including the side whose contact flag is "present", and performs the internal boundary frame correction unit 1 so as to perform the internal boundary frame correction only on these small areas.
Control 61. Similarly, the external boundary frame correction unit 1 selects all of the small areas including the sides whose overflow flag is “present” and performs the external boundary frame correction only on these small areas.
62 is controlled. As a result, as shown in FIG. 8, execution or non-execution of the inner boundary frame correction unit 161 and the outer boundary frame correction unit 162 is determined according to the contents of the contact presence / absence flag and the protrusion presence / absence flag (third means). ).

The internal boundary frame correction unit 161 corrects the image data in the field memory 17 by performing an internal boundary frame correction process on the internal small area designated by the control unit 15 in the page memory 11 with the conventional apparatus. Similarly, the external boundary frame correction unit 162 corrects the image data in the field memory 17 by performing the external boundary frame correction process on the external small area designated by the control unit 15 in the page memory 11 as in the conventional apparatus. To do.

As described above, by correcting the image data in the field memory 17, the written character can be reproduced with high accuracy and the correction process can be performed at high speed. Therefore, character recognition can be performed accurately and at high speed.

When the number of correction areas is large, it is expected that it will be difficult to accurately recognize the image data as character data. Therefore, by detecting the number of correction areas and displaying the number of correction areas when a character recognition result is output, the reliability of character recognition can be used as a measure.

Usually, the recognition result by the character recognition device is often corrected manually, but by referring to this reliability, it is possible to preferentially correct a character with low reliability. The efficiency of correction work can be increased.

[0040]

As described above in detail, in the character recognition apparatus of the present invention, the area formed by the inner boundary frame and the outer boundary frame is divided into a plurality of areas, and the area requiring the correction processing is detected. Since the correction process is selectively performed later, the image data can be corrected promptly as compared with the case where detailed correction such as frame removal is performed on the entire region. Also,
It is possible to recognize a character with high accuracy without lowering the correction accuracy.

In particular, when the area requiring the correction processing is not detected, the character recognition processing is immediately performed without performing the correction processing, so that the processing time can be greatly shortened.

[Brief description of drawings]

FIG. 1 is a block configuration diagram of a character recognition device according to an embodiment of the present invention.

FIG. 2 is a block configuration unit of a flag data generation unit according to the present embodiment.

FIG. 3 is an explanatory diagram of each vertex of an internal boundary.

FIG. 4 is an explanatory diagram of a designated direction at each internal boundary vertex.

FIG. 5 is an explanatory diagram of each vertex of the outer boundary.

FIG. 6 is an explanatory diagram of a designated direction at each outer boundary.

FIG. 7 is a block configuration diagram of a correction unit.

FIG. 8 is an explanatory diagram of a control operation for flag data.

FIG. 9 is a block configuration diagram of a conventional character recognition device.

FIG. 10 is an explanatory diagram of a form image.

FIG. 11 is an explanatory diagram of a projection histogram.

FIG. 12 is an explanatory diagram of internal boundaries.

FIG. 13 is an explanatory diagram of cut out image data.

FIG. 14 is an explanatory diagram of internal boundary frame correction processing.

FIG. 15 is an explanatory diagram of external boundary frame correction processing.

[Explanation of symbols]

 11 page memory 12 frame detection unit 13 boundary frame setting unit 14 flag data generation unit 15 control unit 16 correction unit 17 field memory 18 character cutout unit 19 character recognition unit

Claims (3)

[Claims] 1. A character entry frame detecting means for detecting a character entry frame in image data to be processed, and a boundary for setting an inner boundary frame and an outer boundary frame inside and outside the detected character entry frame, respectively. A frame setting unit, a frame region formed by the inner boundary frame, the outer boundary frame, and the character entry frame is divided into a plurality of small regions, respectively, and flag data representing the presence or absence of image data in each small region is generated. Flag data generation means, correction means for determining a correction area based on the flag data and for correcting image data existing in the correction area, and character data recognition means for recognizing the corrected image data as character data A character recognition device comprising: 2. The flag data generating means includes first means for detecting the presence or absence of image data in each internal small area formed by dividing the internal boundary frame for each side, and the external boundary frame. A second means for detecting the presence / absence of image data in each external small area formed by dividing each of the sides, and a third means for determining the correction area according to each detection result. The character recognition device according to claim 1, wherein: 3. An optical image input device that optically scans a recognition object to generate image data including a character entry frame in which characters are entered, and character recognition that recognizes the characters based on the image data. An optical character reading device having a device, wherein the character recognition device is the character recognition device according to claim 1 or 2.