JPH05210759A - Character recognizing device - Google Patents

Abstract

PURPOSE:To comparatively accurately discriminate the height of a character even when an inputted character image is inclined or characters having different height are mixed by dividing a read character string image into plural areas and discriminating the height of a character string in each area. CONSTITUTION:Character image data for one line are sent from an I/O device 6 to a CPU 1. The CPU 1 stores the obtained character image data in an image preserving RAM 4 and divides the inputted character string into plural areas each of which has a prescribed width or the prescribed number of characters. Then the height of a character string in each divided area is found out and the average value of the height in all the areas is found out while using a RAM 3 for an arithmetic buffer memory. Then the CPU 1 executes unit character segmenting processing based upon a character segmentation control program stored in a ROM 2 in accordance with the obtained average value of the character string height. Character data segmented to a unit character are recognized as a character by a character image recognizing part 5.

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 reading characters using a scanner or the like.

[0002]

2. Description of the Related Art A conventional character recognizing device reads a character per line by using a scanner or the like and cuts out the character for each unit character, and cuts out the unit character and a predetermined reference character. Characters are recognized by comparing patterns. For this character recognition work, it is necessary to determine the height of the cut-out character string (row), and in the conventional character recognition device, it is between the highest point and the lowest point of the cut-out character string image. Is treated as the line height (character string height).

[0003]

However, in the case of simply comparing the highest point and the lowest point to obtain the height of a line, the input character string may be slanted, or the height may be high like a European character string. When characters with different lengths are mixed, the line height and the character height may not correspond to each other, which may hinder the work of cutting out the unit character.

For example, comparing FIG. 5 (a) and FIG. 5 (b), the distance between the highest point and the lowest point as one sentence is different even though the words "book" and "pen" are different. Are different, so they are recognized as strings of different heights.

Further, as shown in FIG. 5 (c), when the input character string image is tilted, as compared with when it is not tilted,
The height of the input character string image is recognized as being large.

In particular, when the character size is not constant and the characters are close to each other, such as an English document printed in a so-called proportional type, the characters are cut out for each unit character. When recognizing, a general method is to extract some cutout candidate points, perform character recognition processing at each cutout candidate point, and determine the cutout candidate point with the best processing result as the cutout position. However, when determining the range for extracting such candidate points, the character width is estimated from the height of the character.

However, if the height of the character is not correctly discriminated as in the above-mentioned example, the cutout candidate point cannot be determined, and it is difficult to obtain the character width accurately.

That is, although the standard character width is estimated from the character height, it is difficult to estimate the standard character width if the character height cannot be accurately extracted. If it is necessary to ensure that the correct cutout position falls within the extraction range, it is necessary to enlarge the extraction range, and many cutout candidate positions are extracted, resulting in poor processing efficiency. On the other hand, if an attempt is made to improve the processing efficiency, the correct clipping position may not fall within the extraction range, and the processing performance will deteriorate.

Further, there is a drawback that it is impossible to accurately recognize homomorphic characters of different sizes.

The present invention has been made in order to overcome the problems of the conventional character recognition device, and is relatively accurate even when the input character string image is tilted or characters of different heights are mixed. An object of the present invention is to provide a character recognition device that can determine the height of a character.

[0011]

A character recognition device according to the present invention is a character recognition device having an image input means for inputting a character string image of at least one line, and the character string image input by the image input means. Is divided into a plurality of regions having a predetermined width or a predetermined number, the height of the character string for each divided region is extracted, and the CPU 1 is provided as a determination unit that determines the height of the input character string from the extracted data. It is characterized by

[0012]

In the character recognition device having the above structure, the CPU 1
Divides the character string image read for each row into a plurality of areas, and determines the height of the character string for each area. In this way, it is possible to realize a character recognition device that is not easily affected by the arrangement and inclination of character strings.

[0013]

Embodiments of the present invention will be described below with reference to the drawings.

FIG. 1 is a block diagram showing the configuration of an embodiment of the character recognition device of the present invention. Data obtained from an input character string reading means such as a scanner (not shown) is transmitted to an input / output device 6 for transmission. The generated character image data for one line is sent to the CPU 1 as a discriminating means via the bus line 7.

The CPU 1 stores the obtained character image data in the image storage RAM 4 and divides the input character string into areas each having a predetermined width or a predetermined number of characters. Next, the height of the character string in each of the divided areas is obtained, and the average value of the heights over all the areas is obtained while using the RAM 3 which is a buffer memory for calculation.

Next, the CPU 1 cuts out the unit character from the obtained average value of the heights of the character strings based on the character cut-out control program stored in the ROM 2.

The character data cut out into unit characters is compared and compared with a reference pattern by the character image recognition unit 5 to perform character recognition.

FIG. 2 is a flow chart for explaining the operation of the character recognition processing performed by the CPU 1 of FIG. 1. First, when a character string image is input from a scanner (not shown), the CPU
1 stores the input image in the RAM 4 in step S1. Next, the CPU 1 divides the character string image into areas of a predetermined width in step S2.

Next, the CPU 1 obtains the height of each area in step S3, calculates the average value of the heights per row, and if the row height is obtained, obtains it in step S4. Cutout processing is performed for each character based on the line height. The cut-out characters are subjected to character recognition processing in step S5.

In step S6, it is determined whether the character recognition processing is completed. If there is a character for which the character recognition processing is not completed, the processing from step S4 is repeated.

If it is determined in step S6 that the character recognition processing has been completed for all the characters, the CPU 1 recognizes in step S7 whether or not the characters of different sizes have the same shape. Whether or not it is completed is determined in step S8. If not completed, the processes from step S3 are repeated.

When the homomorphic character recognition process has been completed, the character recognition process is terminated.

FIG. 3 shows a specific example of the height discrimination of a character string by the character recognition device of the present invention. In FIG. 3, one line character string is divided into, for example, 20 mm widths into five areas. Is divided into The height of the character string is obtained for each area, and the average value is used as the height of the character string.

As a result, in the conventional character recognition device for a character string, if the height of the character string in FIG. 3 (a) is 7 mm, the same character string is tilted and input in FIG. 3 (b). About 2
Although it is as small as 0 mm, the difference between the two is small in the character recognition device of the present invention.

FIG. 4 is a diagram for explaining the operation of the present invention by taking a Roman character printed in proportional typeface as an example, and the height of the character "multi" by the boundary lines A to G in each region. It has been demanded. Since the result of cumulative addition of height direction data in each area is expressed in a histogram, especially the protruding portion in the height direction means that at least a part of the character exists, and the height is low. The part means that it is between characters or a part of a character drawn horizontally.

Therefore, when the characters are close to each other as in the proportional typeface, at least a portion having a low cumulative addition value of heights in each area can be extracted as a cutout candidate point.

[0027]

As described above, according to the character recognition apparatus of the present invention, the CPU 1 divides the input character string image into a plurality of areas and determines the height of the character, so that the character input at an inclination is obtained. It is possible to correctly determine the height of the character string even in the case of a column image, and it is possible to accurately and easily cut out characters and recognize homomorphic characters.

[Brief description of drawings]

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

FIG. 2 is a flowchart showing an operation performed by a CPU 1 in FIG.

FIG. 3 is a diagram showing a specific example of character string height determination performed by the character recognition device of the present invention.

FIG. 4 is a diagram showing a specific example of character cutting processing performed by the character recognition device of the present invention.

FIG. 5 is a diagram showing a specific example of character string height determination by a conventional character recognition device.

[Explanation of symbols]

 1 CPU (discriminating means) 2 ROM 3 RAM for calculation 4 RAM for image storage 5 Character recognition unit 6 Input / output device

Claims (1)

[Claims] 1. A character recognition device having an image input means for inputting a character string image of at least one line, wherein the character string image input by said image input means is divided into a plurality of regions of a predetermined width or a predetermined number. Then, the character recognition device is provided with a determination means for extracting the height of the character string for each divided area and determining the height of the input character string from the extracted data.