JPH09237317A - General document reader - Google Patents

Abstract

PROBLEM TO BE SOLVED: To easily execute the correction of a character recognizing result by improving the recognizing rate of a general document reader. SOLUTION: This reader consists of a scanning part 4 which scans and photoelectricaly converts a character area and the character part of a table area which are obtained by dividing a document picture into the areas of characters, tables, ruled lines and graphics and outputs a picture signal, a character recognizing part 5 setting similarity between the picture and a previously registered standard character shape to be a distance value selecting a character of this distance value within a prescribed range in the order from the small distance value as a candidate character, and an output forming part 6 outputting, in line characters of most small distance values from the recognizing result of this character recognizing part. In this case the character recognizing result is outputted based on an optional character string specified by an operator and the character kind of this character string.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a general document reader for recognizing characters forming a general document, and more particularly to a general document reader for recognizing a specified range with a specified character type.

[0002]

2. Description of the Related Art Generally, in a conventional general document reading apparatus,
The document image to be read is automatically or manually divided into character, table, ruled line, figure, and photograph areas, and character recognition is performed on the character portions of the character area and table area. In character recognition, a standard character shape, which is a character pattern registered in a recognition dictionary, is compared with image data cut out as a character, and a character recognition result is selected. In the recognition dictionary, for example, in the case of a dictionary for reading Japanese sentences, kanji, hiragana, katakana, numbers, letters, symbols, etc. are registered.

[0003]

In the character recognition of the conventional general document reading apparatus, for example, the lowercase letter "l" and the numeral "1", the alphabetical letters "o", "O" and the numeral "0", and the katakana character "me" are used. "" And the symbol "x" and the like have different character types but very similar glyphs and are often misrecognized. Therefore, a result such as a number being mixed in the alphabetic character string will occur.
In this case, since the correct character and the character resulting from the erroneous recognition are similar to each other in shape, there is a problem that it is easy to overlook an error during correction and it is difficult to correct.

As described below, the present invention improves the recognition rate and simplifies the correction of the recognition result. The recognition rate is improved by recognizing the character string designated by the operator with the character type designated by the operator at a portion recognized as a character by the general document reading device. When correcting the character recognition result of a general document reading device, regarding the character recognition result, the character string specified by the operator
By recognizing again with the character type specified by the operator, it becomes easy to correct the character recognition result. When correcting the character recognition result of the general document reading device, for the character recognition result, the operator specifies the range and the character type, and by selecting only the candidate character of the specified character type from the candidate characters at each character position of the range, It facilitates the correction of character recognition results.

[0005]

SUMMARY OF THE INVENTION According to the present invention, a document image is divided into regions such as characters, tables, ruled lines, figures and photographs, and the character portions of the character region and the table region are scanned and photoelectrically converted to obtain an image. A scanning unit that outputs a signal and a degree of similarity between the image and a standard character shape registered in advance are used as distance values, and characters having a distance value within a predetermined range are selected as candidate characters in order from the smallest distance value. In a general document reading device including a character recognition unit and an output forming unit that arranges and outputs the character having the smallest distance value from the recognition result of the character recognition unit, an arbitrary character string designated by an operator and the character string The feature is that the character recognition result is output based on the character type information.

[0006]

Embodiments of the present invention will be described below with reference to the drawings. [First Embodiment] FIG. 1 is a block diagram showing the configuration of the first embodiment. In the figure, reference numeral 1 indicates a form to be read. Reference numeral 2 denotes an image input device, which is used to capture the form 1 as a document image. A layout analysis unit 3 automatically or manually divides the document image into areas such as characters, tables, ruled lines, figures, and photographs.

A scanning unit 4 scans the area determined by the layout analysis unit 3 to be a character area and the character portion in the table area, and an image signal obtained by photoelectric conversion is scanned by the character recognition unit 5.
Transfer to The character recognition unit 5 forms a set of candidate characters for the input character image and their distance values, and outputs the set to the output forming unit 6. The standard character shape for selecting the candidate character is registered in a recognition dictionary (not shown).

The output forming unit 6 is for arranging and outputting the character having the smallest distance value from the recognition result of the character recognition unit 5. Next, the processing procedure of the general document reading apparatus having the above configuration will be described using a specific recognition target. First, the document image input from the image input device 2 is processed by the layout analysis unit 3
, It is automatically or manually divided into character, table, ruled line, figure and photo areas. FIG. 2 is an explanatory diagram showing an example of a region recognized as a character.

The scanning section 4 scans the area determined by the layout analysis section 3 as a character area and the character portion in the front area, and transfers an image signal obtained by photoelectric conversion to the character recognition section 5. The character recognition unit 5 calculates the distance between the glyph of the input character and the standard glyph of each character, forms a set of candidate characters and distance values arranged in ascending order of distance (that is, in the order of similar glyphs), and outputs the result. Output to the forming unit. Standard glyphs are registered in the recognition dictionary. In the recognition dictionary, for example, in the case of a Japanese sentence reading dictionary, kanji, hiragana, katakana, numbers, letters, symbols, etc. are registered. FIG. 3 shows an explanatory diagram of a recognition target and a character recognition example.

The output forming unit 6 arranges and outputs the character having the smallest distance value from the recognition result of the character recognition unit 5. FIG.
Is an explanatory diagram showing an output example of erroneous recognition. In the example of this figure, "o" (lowercase letters) and "0" (numeric characters) have different character types, but it is difficult to determine which is correct. Therefore, the judgment at the time of character recognition is prioritized. FIG.
Is an explanatory diagram showing an output example of erroneous recognition, and the conventional device will output a result as shown in this figure.

According to the present embodiment, the operator designates an image of an arbitrary character string in which a recognition error is considered to occur in a portion determined to be a character by the layout analysis, and at that time, the character string Characteristic is also specified, and the recognition rate is improved by recognizing the character string only with the specified character type.

FIG. 5 is an explanatory diagram showing an example of character string / character type designation for a document image, and the character recognition of the present embodiment will be further described with reference to this figure. First, an image of an arbitrary character string is designated on the image display of the portion determined to be the character region, and the character type of the image of the character string is also designated. In the example of FIG. 5, in the specified area, alphabetic characters and
The numbers and katakana are specified.

Next, character recognition is performed based on the designation of FIG. For the image of the specified character string, character recognition is performed only with the standard glyph of the specified character type. For character image portions other than those designated, character recognition is performed using all standard glyphs registered in the dictionary. FIG. 6 is an explanatory diagram showing an example of character recognition according to the present embodiment, which is an example of performing character recognition based on the designation.

FIG. 7 is an explanatory diagram showing an output example of the embodiment, and is an example in which an output character string is selected from the character recognition result shown in FIG. Before recognizing a character, a character type is specified and recognized for an arbitrary character string, so that a recognition error can be prevented. As described above, in the present embodiment, the recognition rate can be improved by recognizing the image of the character string designated by the operator with the character type designated by the operator. [Second Embodiment] In the present embodiment, in the configuration of the first embodiment, an operator can specify a character string for a portion where a character recognition result error occurs after the character recognition result, and the specification is performed. The feature is that the specified character string is recognized again with the specified character type.

First, layout analysis is performed on the document image input from the image input device 2 to obtain the document image of FIG. For this image, the scanning unit 4 and the character recognition unit 5
Is used for character recognition, and the character recognition result of FIG. 3 is obtained.
The output forming unit 6 arranges and outputs the character having the smallest distance value from the recognition result of the character recognition unit 5. For example, an output example as shown in FIG. 4 is obtained. For example, "o" (lowercase letters) and "0" (numbers) have different character types, but it is difficult to determine which is correct. As a result, the judgment at the time of character recognition is prioritized as in the example of FIG.

In the present embodiment, the operator designates an arbitrary character string whose character types are to be aligned as a result of character recognition of a portion judged to be a character by layout analysis, and at that time, the character string The character recognition result can be easily corrected by designating the character types together and recognizing the character string again with the designated character type.

FIG. 8 is an explanatory view showing an example of character string / character type designation for a character recognition result. The character recognition of this embodiment will be described below. First, an arbitrary character string is specified on the character recognition result shown in FIG. 4, and the character type of the character string is also specified. FIG. 8 shows the character recognition result of FIG.
This is an example in which an arbitrary character string is specified and alphabetic characters, numbers, and katakana are specified.

Next, character recognition is executed again based on the designation in FIG. Only for the character string designated in FIG. 8, character recognition is performed again with the standard character shape of the designated character type, and character recognition other than the designated character is not performed again. FIG. 9 is an explanatory diagram showing the character recognition result of the present embodiment, which is an example in which character recognition is performed again based on the designation in FIG. As a result, the correct output result as shown in FIG. 7 can be obtained.

As described above, in the present embodiment, the operator designates a character string for the portion where the character recognition result error occurs after the character recognition result, and at the same time, character recognition is performed again by the designated character type. It is possible to save the trouble of correcting each character by one character and easily correct the character recognition result. [Third Embodiment] In the present embodiment, in the configuration of the first embodiment, an operator can designate a character string for a portion where a character recognition result error occurs after the character recognition result, and the designation is performed. The feature is that only candidate characters of the specified character type are selected for the specified character string.

First, the document image input from the image input device 2 is subjected to layout analysis to obtain the document image shown in FIG. For this image, the scanning unit 4 and the character recognition unit 5
Is used for character recognition, and the character recognition result of FIG. 3 is obtained.
The output forming unit 6 arranges and outputs the character having the smallest distance value from the recognition result of the character recognition unit 5. For example, an output example as shown in FIG. 4 is obtained. "O" (lowercase letters) and "0"
(Number) has different character types, but it is difficult to determine which is correct. As a result, the judgment at the time of character recognition is prioritized as in the example of FIG.

In the present embodiment, the operator designates an arbitrary character string whose character types are to be aligned as a result of character recognition of a portion determined to be a character by layout analysis, and at that time, the character string It is possible to easily correct the character recognition result by designating the character types together, selecting only the candidate characters of the designated character type for the character string, and displaying them again as the character recognition result.

Character recognition according to this embodiment will be described below. First, an arbitrary character string is specified on the character recognition result shown in FIG. 4, and the character type of the character string is also specified.
FIG. 8 is an example in which an arbitrary character string is specified on the character recognition result of FIG. 4, and alphabetic characters, numbers, and katakana are respectively specified. Next, based on the designation in FIG. 8, only candidate characters that fit the designated character type are selected and newly output as a candidate character group. Characters other than the designated characters are not selected as candidate characters by character type.

FIG. 10 is an explanatory diagram of the candidate character selection according to the present embodiment, and shows an example in which the candidate character selection is performed based on the designation of FIG. As a result, the correct output result as shown in FIG. 7 can be obtained. As described above, in the present embodiment, with respect to the portion where the character recognition result error occurs after the character recognition result, the operator designates the character string, and at the same time, selects only the candidate characters of the designated character type,
The operator is not required to correct each character, and the character recognition result can be easily corrected.

In the first to third embodiments described above, the designation of kanji, hiragana, katakana, numbers, letters, symbols, etc. has been described as an example, but the present invention is not limited to this. The present invention can be similarly applied to the recognition of documents in the languages and script systems. Furthermore, uppercase letters and lowercase letters can be specified in more detail for English letters. For example, all uppercase letters, all lowercase letters, and the first letter can be specified as uppercase letters and lowercase letters.

Further, although it has been described that a document image is obtained by reading a form or the like, the present invention is not limited to this, and the present invention can be similarly applied to a device for recognizing handwritten characters by pen input. is there.

[0026]

As described in detail above, the operator designates the character type of an arbitrary object and outputs the character recognition result based on this designation, whereby the recognition rate can be improved, and the character recognition can be improved. This has an effect that the result can be easily corrected.

[Brief description of drawings]

FIG. 1 is a block diagram illustrating a configuration of an embodiment.

FIG. 2 is an explanatory diagram showing an example of a region recognized as a character.

FIG. 3 is an explanatory diagram of a recognition target and a character recognition example.

FIG. 4 is an explanatory diagram showing an output example of erroneous recognition.

FIG. 5 is an explanatory diagram showing an example of specifying a character string / character type for a document image.

FIG. 6 is an explanatory diagram showing an example of character recognition according to the first embodiment.

FIG. 7 is an explanatory diagram showing an output example of the embodiment.

FIG. 8 is an explanatory diagram showing an example of character string / character type designation for a character recognition result.

FIG. 9 is an explanatory diagram showing a result of character recognition according to the second embodiment.

FIG. 10 is an explanatory diagram of selecting a candidate character according to the third embodiment.

[Explanation of symbols]

 4 Scanning part 5 Character recognition part 6 Output forming part

Claims (3)

[Claims] 1. A scanning unit that divides a document image into areas such as characters, tables, ruled lines, figures, and photographs, and scans the character portions of the character areas and table areas to perform photoelectric conversion and output an image signal. , A character recognition unit that sequentially selects, as a candidate character, a character whose distance value is within a predetermined range with a degree of similarity between the image and a standard character shape registered in advance as a candidate character, and the character. In a general document reading device, which comprises an output forming unit that arranges and outputs the character having the smallest distance value from the recognition result of the recognition unit, the character recognition unit is configured to display an image of an arbitrary character string of a character portion designated by an operator and Based on the character type information of the image of the character string, character recognition is performed for any range specified by the operator using only the standard glyphs of the specified character type, and the range other than the specified range is registered. General document reading device being characterized in that shall perform character recognition using the entire quasi-shape. 2. A scanning unit that divides a document image into areas such as characters, tables, ruled lines, figures, and photographs, and scans the character portions of the character areas and table areas to perform photoelectric conversion and output an image signal. , A character recognition unit that sequentially selects, as a candidate character, a character whose distance value is within a predetermined range with the degree of similarity between the image and a standard character shape registered in advance as a distance value, and the character. In a general document reading device consisting of an output forming unit that arranges and outputs the character with the smallest distance value from the recognition result of the recognizing unit, the character recognizing unit is used for the output result from the output forming unit The general document reading device is characterized in that the character string of is recognized again with the designated character type. 3. A scanning unit that divides a document image into areas such as characters, tables, ruled lines, figures and photographs, and scans the character portions of the character areas and table areas to perform photoelectric conversion and output an image signal. , A character recognition unit that sequentially selects, as a candidate character, a character whose distance value is within a predetermined range with a degree of similarity between the image and a standard character shape registered in advance as a candidate character, and the character. In a general document reading device including an output forming unit that arranges and outputs the character having the smallest distance value from the recognition result of the recognition unit, the output forming unit is designated by the operator with respect to the output result from the output forming unit. A general document reading device characterized in that, for an arbitrary character string, only candidate characters of a designated character type are selected and output again.