JP2000315247A - Character recognizing device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a character-recognizing device capable of recognizing characters exactly and at high speed even from a character image the characters of which are difficult to be recognized. SOLUTION: This character recognizing device is provided with a character area segmenting part 11 to estimate character width of the inputted character image 10 and to segment a character area with length equivalent to the character width, a feature extracting part 12 to extract features of a character pattern in the segmented character area, a standard pattern dictionary 14 consisting of both of a one-character standard pattern dictionary 14-1 and a plural character standard pattern dictionary 14-2 and a dictionary collating part 15 to extract candidates for the character or character strings to be expressed by the character pattern in the character area by collating features of the character pattern extracted by the feature extracting part 12 with features of each of one character and the character strings described in the one-character standard pattern dictionary 14-1 and the plural character standard pattern dictionary 14-2.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

The present invention relates to a character recognition device for recognizing an input character image as a character.

[0002]

2. Description of the Related Art Conventionally, in the field of character recognition devices for recognizing input character images as characters, for example, OCR
(Optical character reading device) and a character recognition device combining an image scanner and OCR software are used in a wide range of business fields. In the conventional OCR form, a form in which a character entry box is printed for each character is used, and in a system using kanji, a large character entry box is used for the size of the character entry box so that kanji can be easily entered. This makes it easier for the OCR to accurately recognize the entered characters, and has the effect of encouraging the writer to enter the characters without touching the adjacent characters.

However, as the field of use of OCR becomes more widespread, a character string written using a small form such as a general form used without being bound by a character frame, that is, an irregular form such as a normal form is used. There is a need for a character recognition device capable of recognizing a low-quality handwritten character string in which adjacent characters written by an irregular writing method are in contact with each other with practical recognition accuracy. I have.

In the case of a printed character string, an OCR
-OCR forms carefully printed using only a plurality of specific character fonts, such as B or multi-font, are used.
Character strings with various deformations printed by unspecified printers at various character pitches, for example, character strings printed at narrow character pitch printers or proportional pitches, or character strings containing half-width and full-width characters mixed For example, there is a demand for a character recognition device that can cut out a character string with a recognition accuracy that is practical.

[0005]

In the character extraction system in the conventional character recognition apparatus, particularly in the character extraction system from a free-pitch character string, that is, a field in which a character frame is not set, a connected area of isolated black pixels is used. A method is employed in which height information of a character string is obtained, pitch information of the character is obtained from the height information, and adjacent characters are integrated or separated based on the value to cut out one character to be recognized. Alternatively, a character string of one line and one field is projected in a direction perpendicular to the line direction and the field direction,
There is also a method in which character pitch information is obtained from character string height information, and character positions are cut out from a projection histogram based on the value. If the cut-out is not determined in one way, a plurality of cut-out positions may be obtained.

Normally, characters written by an unspecified number of writers in accordance with a predetermined format are collected, and from these characters, the characteristics of the character pattern depending on the recognition method are extracted. Is created.
Further, the standard pattern may be created as an average pattern from the collected characters. In the case of handwritten characters, a large character deformation is caused by the writer, so that a plurality of standard patterns are created for each character type. Usually, one standard pattern is called a template, and a dictionary in which a plurality of standard patterns are collected is called a multiple template dictionary. As a dictionary matching method, a feature of a character pattern is extracted from one character cut out from an input form, similarity calculation or distance calculation is performed between this feature and a feature of a template of a standard pattern dictionary, and the maximum similarity is calculated. Or a category of a template having a minimum distance as a recognition result. As candidate character strings, candidate character strings up to the eighth place, for example, are set in ascending order of distance.

[0007] For print characters, a candidate character string is obtained in a similar manner. When there are a plurality of candidates, a method of determining the best clipping candidate by the similarity evaluation value between patterns or by language processing such as knowledge processing is adopted.

In general, when recognizing a character string written at a free pitch or a character string printed by an unspecified printer as in the case of writing by hand on a normal slip,
Adjacent characters often touch each other, and in the case of a print character string, half-width characters and full-width characters may be mixed, so that adjacent characters often touch each other, so that each character is detected and recognized one by one. According to the method, it is difficult to determine which range is a single character, and it is difficult to realize a recognition accuracy that can withstand practical use. For hiragana, katakana, and kanji characters, there are characters that originally consist of separated character areas, such as “n” and “n”, and if these character areas cannot be recognized correctly, it is possible to determine how many characters have been entered. There are times when you can't.

For character extraction of a print character string, the following various methods have been proposed as methods for estimating the pitch of each character from a character string of one line.

In the first method, black pixels of a character string in one line are projected in the line direction to determine the average height of characters forming the line.
Next, the black pixels of the character string are projected in a direction perpendicular to the row direction, and the frequency of the black pixels in the row direction is represented as a histogram. , Ie, a cutting candidate position at which a contact character or an overlapping character is to be separated. Based on this cutting candidate position, a position that is considered to be valid as a full-width character and half-width character string based on the average height of the character string previously obtained is set as a character cutout position.

The second method is to project the black pixels of a character string in one line in a direction perpendicular to the line direction and to represent the frequency of the black pixels in the line direction as a histogram. Is obtained, and a plurality of positions estimated as character extraction positions are obtained for the character string. The determination of which of the plurality of character cutout positions is a likely character cutout position is left to the judgment based on the similarity evaluation value of the character string and the recognition result.

In a third method, a connected area of black pixels is obtained for one line of a character string, small areas are integrated with respect to the arrangement of the connected areas, and a large area is divided into a plurality of areas to determine one character area. Cut out. At this time, the small area and the large area are determined based on the height of the character string.

However, each of the above methods has the following problems. That is, in the first method, when the character string of one line is composed entirely of full-width characters, the full-width characters are considered to have the same character width and character height. Can be accurately estimated, and thus can be cut out exactly one character at a time.However, if half-width characters and full-width characters are mixed, for example, half-width alphanumeric characters or katakana are mixed in full-width kana mixed kanji character strings In such a case, the character position may be shifted by one half-width, or two half-width characters may be recognized as one character, and a correct character extraction position may not be obtained.

In the second method, a plurality of cut-out results are obtained from a one-line character string, so that it takes a long processing time for a one-line recognition process for a plurality of lines, and there are many contact characters. However, there is a problem that the recognition processing speed is significantly reduced because a large number of cutout results are obtained.

In the third method, there are many problems in recognition accuracy regarding separation of a touched character string portion, and it is difficult to perform accurate character recognition.

The present invention has been made in view of the above circumstances, and has as its object to provide a character recognition apparatus capable of accurately and quickly recognizing characters even from a character image in which character recognition is difficult.

[0017]

A character recognition apparatus according to the present invention for achieving the above object estimates a character width from an input character image and calculates a character area having a length corresponding to the character width from the character image. A character area cutout section to be cut out, a feature extraction section to extract the characteristics of the character pattern in the character area cut out by the above character area cutout section, and a one-character standard pattern dictionary in which the characteristics of each standard character are described. And a standard pattern dictionary composed of both a character pattern in which a predetermined number of standard characters are arranged and a multi-character standard pattern dictionary in which the characteristics of each character string are described. The features of the pattern are the features of each of the one-character and character strings described in the one-character standard pattern dictionary and the multi-character standard pattern dictionary that constitute the standard pattern dictionary. Combined and is characterized in that a dictionary collation unit for extracting a candidate character or character string representing a character pattern of the character area.

Here, the character recognition device of the present invention compares the character pattern feature in the character area with the character string feature described in the multi-character standard pattern dictionary by the dictionary matching unit. When the candidate of is extracted,
The feature extraction unit causes the character region in which the character string candidate is extracted to further extract the characteristic of the character pattern of each divided region obtained by dividing the character region into the same number as the number of characters constituting the character string. The feature of each character pattern of the divided area is compared with the feature of each character described in the one-character standard pattern dictionary to extract a character string candidate corresponding to the arrangement of the divided area, and the character string candidate Is preferably provided with a verification unit for verifying whether or not it matches a candidate for a character string extracted by using the plural-character standard pattern dictionary.

Further, the character area cutout section may divide a character area having a length corresponding to the character width from the character image by a length corresponding to one division when the character width is divided into the same number as the predetermined number. It is also preferable to cut out while shifting each other.

The dictionary collating unit extracts a character or character string candidate for each of the character regions cut out by the character region extracting unit, and synthesizes the extracted character or character string candidates to form a character arrangement. It is also a preferable embodiment to extract the candidate of (1).

Further, the dictionary collating unit, when a character unrecognizable area in which the character cannot be recognized remains, identifies the characteristics of the character pattern in the character unrecognizable area with the characteristics of each character described in the one-character standard pattern dictionary. Extracting character candidates by collation is also a preferred embodiment.

[0022]

Embodiments of the present invention will be described below.

FIG. 1 is an external view (a) and an internal configuration diagram (b) of a computer on which the character recognition device of the present invention is operated.
It is.

As shown in FIGS. 1A and 1B, the computer 60 has a CPU 61 and a memory 6.
2. Display device 63, printer 64, CD-ROM drive 65, magnetic disk 66, mouse 67, keyboard 6
8 and an image scanner 69 are connected to the system bus 7.
The character recognition device of the present invention is formed on the computer 60.

A character image is input to the computer 60 by the image scanner 69, and character recognition is performed.

FIG. 2 is a functional block diagram of the character recognition device according to the first embodiment of the present invention.

As shown in FIG. 2, the character recognition apparatus 100 according to the first embodiment estimates a character width from an input character image 10 and calculates a character area having a length corresponding to the character width from the character image 10. A character region extracting unit 11 for extracting characters, a character extracting unit 12 for extracting characteristics of a character pattern in a character region extracted by the character region extracting unit 11, a one-character standard pattern in which the characteristics of each standard character are described. Dictionary 1
A multi-character standard pattern dictionary 1 that describes the characteristics of each character string in which a predetermined number of 4_1 and standard characters are arranged
4_2, the one-character standard pattern dictionary 14_1 and the plural-character standard pattern dictionary 14 constituting the standard pattern dictionary 14 are extracted from the character pattern in the character area extracted by the feature extracting unit 12.
_ 2, a dictionary matching unit 13 that matches a feature of each character and character string described in _ <b> 2 to extract a candidate for a character or a character string represented by a character pattern in the character area, and a candidate for the extracted character or character string. A candidate character output unit 15 for storing and outputting to the outside is provided.

Next, the operation of the character recognition device 100 according to the first embodiment will be described.

For example, when a character image 10 fetched from a form or the like by an image scanner or the like is input to the character recognition device 100, an average character height is obtained by the character area cutout unit 11, and the character height is calculated based on the average height. The width is estimated. The average height of a character is determined by projecting the black pixels of the character string in the row direction, and based on the assumption that the width and height of Japanese double-byte characters are often written almost equally. From this, the character pitch, that is, the character width can be estimated. Full-width characters can be detected based on the character width.

Since a half-width character has half the width of a full-width character, one half-width character can be detected with a half width of the average character height.

For a contact character in which adjacent characters are in contact with each other, if the width of the character area is two characters of the above character width, the character area is forcibly divided into half. Thus, it can be detected as two characters. If there is a width of three characters, the character area may be forcibly divided into one third.

A character area having a length corresponding to the character width thus obtained is sequentially cut out from the left end of the character image 10.
Estimation of the character width and extraction of the character area are performed using the character image 10
May be configured not only from the left end, but also from the middle of the character image 10 as necessary.

The feature extracting unit 12 extracts the features of the character pattern in the cut-out character region. The feature of the character pattern is obtained, for example, by dividing a rectangular area for one character into 8 × 8 = 64 areas of 8 columns × 8 rows and scanning in the vertical and horizontal directions to count the number of black pixels. Can be represented by the following projected distribution function: The feature of the character pattern in the character region extracted by such a method is compared in the dictionary matching unit 13 with the feature of each of the one character and the character string described in the one-character standard pattern dictionary 14_1 and the plural-character standard pattern dictionary 14_2. For example, ten candidates are obtained in ascending order of the distance by obtaining the distance between them.

The standard pattern dictionary 14 (see FIG. 2) includes a one-character standard pattern dictionary 14_1 statistically created from a normal one-character character area and a predetermined number of standard characters, for example, two. And a multi-character standard pattern dictionary 14_2 in which the characteristics of each character string are described.

As described above, the two types of dictionaries, the one-character standard pattern dictionary 14_1 and the plural-character standard pattern dictionary 14_2, are provided as the standard pattern dictionary 14 in this embodiment because half-width numbers are arranged in a full-width character string. In this case, the character area cutout unit 11 detects two half-width characters as one character, so that dictionary matching is performed using the two half-width characters as one pattern.

The standard pattern dictionary 14 describes standard patterns and their categories, that is, character codes representing the character categories in which the standard patterns are created. In the case of a one-character standard pattern dictionary, one character code is described, for example, "0". In the case of a multiple-character standard pattern dictionary, a sequence of two character codes is described, for example, "00". ing.

The result obtained by the collation using the standard pattern dictionary of one character in the dictionary collation unit 13 is determined to be a character area that can be detected as a normal one character, so that one character code is output as a candidate as it is. I do. The result obtained by the collation using the plural-character standard pattern dictionary in the dictionary collating unit 13 outputs a sequence of a plural-character code as a candidate.

In the candidate character output unit 15, the dictionary matching unit 13
A final candidate character string is created by organizing the character codes output from, and a plurality of character candidates are output to the outside as a sequence of a plurality of character codes while one character candidate remains one character code.

As described above, the character recognition device 1 of the present embodiment
In the case of 00, half-width characters are included by focusing on a local character string portion instead of the entire character line and performing matching using a standard pattern dictionary composed of both a one-character standard pattern dictionary and a multi-character standard pattern dictionary. Characters can be cut out accurately even from low-quality character strings where characters touch each other, increasing the accuracy of character recognition and eliminating the need to extract many cutout candidates. Processing time is also reduced.

Next, a second embodiment of the character recognition device of the present invention will be described.

FIG. 3 is a schematic configuration diagram showing a second embodiment of the character recognition device of the present invention.

As shown in FIG. 3, a character recognition apparatus 200 according to the second embodiment has the same character area extraction unit 11 and feature extraction as the character recognition apparatus 100 according to the first embodiment shown in FIG. Unit 12, a standard pattern dictionary 14, which includes both a single-character standard pattern dictionary 14_1 and a multiple-character standard pattern dictionary 14_2, a dictionary matching unit 13, and a candidate character output unit 15, and a verification unit 16 is added. .

The verification unit 16 compares the characteristics of the character patterns in the character area with the characteristics of the character strings described in the multiple-character standard pattern dictionary 14_2 by the dictionary matching unit 13, and extracts character string candidates. In this case, the feature extraction unit 1
2, the character region in which the character string candidate is extracted is further extracted with the same number of characters as the number of characters constituting the character string. The character pattern feature of each region is compared with the feature of each character described in the one-character standard pattern dictionary 14_1 to extract a character string candidate corresponding to the arrangement of the divided regions. This is to verify whether or not it matches the character string candidate extracted using the standard pattern dictionary 14_2.

FIG. 4 is a flowchart for explaining the operation of the verification unit provided in the character recognition device of the second embodiment.
FIG. 9 is an explanatory diagram of a verification function according to the second embodiment.

As shown in FIG. 4, it is determined whether the recognition result is a recognition result of a plurality of character patterns (step S).
1) If the result of the determination is a recognition result of a plurality of character patterns, that is, the dictionary matching unit 13 matches the features of the character patterns in the character area with the features of the character strings described in the multiple-character standard pattern dictionary 14_2. As a result, when a character string candidate is extracted, the character area is cut out again (step S2). That is, the feature extracting unit 12 causes the character region in which the character string candidate is extracted to further extract the character pattern characteristics of each divided region obtained by dividing the character region into the same number as the number of characters constituting the character string.

The process up to this point will be described with reference to FIG. 5.
A plurality of characters, for example, two characters “0” are one pattern 2
If it is recognized as 0, the character region is cut out by the character region cutout unit 11, and division is performed at a position indicating the minimum value of the black pixel projection histogram in the vertical direction of the pattern 20. In this example, there are two characters, so pattern 20_1
And a pattern 20_2. If two half-width characters are arranged, they are divided into two equal parts.

Next, each of the divided characters is recognized (step S3). That is, the dictionary matching unit 13
Are the divided character patterns 20_1 and 20_1.
_2 is compared with the characteristics of each character described in the one-character standard pattern dictionary 14_1 to extract character string candidates corresponding to the arrangement of the divided regions.

Next, verification of the recognition result is performed (step S4). That is, the verification unit 16 verifies whether or not the character string candidate extracted using the one-character standard pattern dictionary 14_1 matches the character string candidate extracted using the multiple-character standard pattern dictionary 14_2.

The result of the verification by the verification unit 16 (step S
5) A character string candidate extracted using the one-character standard pattern dictionary 14_1 is a plural-character standard pattern dictionary 14_2.
If the character string matches the candidate for the character string extracted using
Character codes 21 and 22 of 0_1 and 20_2 (see FIG. 5)
Is output to the candidate character output unit 15. If the result of determination in step S5 is that they do not match, the division is determined to be inappropriate and the division result is excluded from the candidates (step S5).
6).

In the second embodiment, since such verification is performed, accurate character recognition can be performed even from a character image when two characters come into contact with each other as shown in FIG.

Next, a third embodiment of the character recognition device of the present invention will be described.

FIG. 6 is an explanatory diagram of a character cutout method in a third embodiment of the character recognition device of the present invention.

The character recognition device according to the third embodiment has a configuration similar to that of the character recognition device according to the first or second embodiment shown in FIG. 2 or FIG. Only the function is different from the character recognition devices of the first and second embodiments. That is, in the character area cutout unit of the character recognition device according to the third embodiment, the character area having the length corresponding to the character width is divided by two when the character width is divided into two from the input character image. It is configured to cut out while shifting by the length of.

A case where a character area is cut out from a character image 30 corresponding to ten full-width character strings from "1" to "0" as shown in FIG. 6 will be described.

First, in the case of the character image 30 which is estimated to be a horizontally written character string, the black pixels of the character image 30 are projected in the vertical direction to create a histogram. Find the minimum position,
Judged from the height H, H and H / 2 from the top of the character image
W and W / 2 are estimated from the interval of the position of the minimum value appearing at a distance corresponding to. And the pitch is W by this character width W.
The character area is cut out while shifting by / 2.

In the case of FIG. 6, first, the character area "12" of al
Is extracted as one character, feature extraction and dictionary matching are performed, and candidates are obtained. In the case of this character area “12”, the distance is closest to the standard pattern of “12” in the multiple-character standard pattern dictionary 14_2 (see FIG. 3), so the character is cut again in the same manner as in the second embodiment. Output, feature extraction, dictionary matching, and verification are performed, and the character area "12"
Is divided into two character regions, and two character candidates of “1” and “2” are obtained. Next, the character area "23" of a2 shifted by W / 2 is cut out as one character, and the same recognition processing as in the case of the character area al is performed.
And "3" are obtained. Similarly, the character areas a3, a4,.
Two character candidates "3" and "4", two character candidates "4" and "5", and so on are obtained.

The dictionary matching unit 13 extracts a character or character string candidate for each of the character regions cut out as described above, and extracts a character arrangement candidate by integrating the extracted character or character string candidates. And outputs it to the candidate character output unit.

By performing character extraction in this manner, accurate character recognition can be performed even for a character image in which full-width characters and half-width characters are mixed.

Further, according to this character extracting method, accurate character recognition can be performed even in the case of the following contact characters.

FIG. 7 is a diagram showing another example of processing by the character cutout method in the third embodiment.

As shown in FIG. 7, the character image 40 includes one "1" and seven "0" s that are in contact with each other. In the case of the character image 40 as well, as in FIG. 6, first, the character area "10" of al is cut out as one character, and feature extraction and dictionary matching are performed to obtain candidates. In the case of the character area “10”, the distance is closest to the standard pattern of “10” in the multiple-character standard pattern dictionary 14_2 (see FIG. 3), and thus “1” is obtained by the same method as in the second embodiment. And two characters “0” are obtained. Next, the character area “00” of a2 shifted by W / 2 is cut out as one character, and two character candidates “0” and “0” are obtained. Similarly,
The areas a3, a4,... are processed.

The dictionary matching unit 13 extracts a character or character string candidate for each of the character regions cut out as described above, and extracts a character arrangement candidate by integrating the extracted character or character string candidates. And outputs it to the candidate character output unit.

By performing character extraction in this manner, accurate character recognition can be performed even for a character image in which adjacent characters contact each other.

Next, a fourth embodiment of the character recognition device of the present invention will be described.

The character recognition device according to the fourth embodiment has substantially the same configuration as the character recognition devices 100 and 200 shown in FIG. 2 or FIG. When a character unrecognizable area remains, the feature of the character pattern is matched with the feature of each character described in the one-character standard pattern dictionary 14_1 to extract a character candidate.

FIG. 8 is an explanatory diagram of a character segmentation method in a fourth embodiment of the character recognition device of the present invention.

As shown in FIG. 8A, in the case of a character image 50 in which a half-width character or a white space is mixed with a character width corresponding to a half-width in the arrangement of full-width katakana characters, as shown in FIG. First, the character width W is estimated from the average height H by the same method as in the second and third embodiments, and the character area al “f” is cut out from the left end of the character image 50 by the character width W to form katakana. Is recognized. Next, W /
The recognition process is performed by shifting the cutout position by 2 and cutting out the character area a2 with the character width W. However, this character area a2
Does not match the dictionary, so no character candidate is obtained.
In the same manner, the processing of cutting out and recognizing each character area of a3, a4, a5, and a6 is continued.

In the case of FIG. 8B, the character area al, the character area a3, and the character area a6 are recognized as katakana character "ka", katakana character "ka", and katakana character "ku", respectively. Area a2, character area a
4 and the character area a5 are rejected as a result of dictionary collation. The character candidates obtained as a result of shifting the cutout position by W / 2 in this manner are integrated by the dictionary matching unit of the present embodiment, and are extracted as character array candidates. In the example of FIG. 8, the character “f” obtained from the character area a1 and the character area a
The character array "Kaka" obtained by integrating the character "K" obtained from the character region a3 and the character "K" obtained from the character region a6 are obtained. A character unrecognizable area a7 where characters cannot be recognized remains.

Therefore, the dictionary collating unit of the present embodiment uses the one-character standard pattern dictionary 14_1 ( (See FIGS. 2 and 2) The character unrecognizable area a7 is recognized as a half-size cloud point by comparing character features of each character and extracting character candidates again. Thus, a recognition result as shown in FIG. 8 (c) is obtained from the character string shown in FIG. 8 (b).

FIG. 9 is a flowchart showing the processing of a character unrecognizable area in the fourth embodiment.

As shown in FIG. 9, the recognition results of the character areas al to an are obtained (step S11).
It is determined whether or not there is an unrecognizable area in n (step S12). If there is an unrecognizable area, the unrecognizable area is regarded as one character, and a dictionary is referenced using the one-character standard pattern dictionary 14_1 to perform a character candidate. Ask for.

As described above, in the fourth embodiment, even if one half-width character is mixed between full-width characters, it can be accurately recognized. Also, as shown in FIG. 8A, a case where a half-width space is mixed between full-width characters can be accurately recognized in the same manner as in the case of the above half-width turbid point.

[0073]

As described above, according to the character recognition apparatus of the present invention, a one-character standard pattern dictionary in which the characteristics of each standard character are described and a character in which a predetermined number of standard characters are arranged Character recognition is performed using a standard pattern dictionary consisting of both a multi-character standard pattern dictionary in which the characteristics of each column are described. It is possible to cut out and recognize characters with high precision even for character strings and print character strings in which full-width characters and half-width characters are mixed, enabling accurate and high-speed character recognition even from character images that are difficult to recognize. Character recognition device can be realized.

[Brief description of the drawings]

FIG. 1A is an external view of a computer in which a character recognition device of the present invention is operated, and FIG.

FIG. 2 is a functional block diagram of the character recognition device according to the first embodiment of the present invention.

FIG. 3 is a schematic configuration diagram showing a second embodiment of the character recognition device of the present invention.

FIG. 4 is a flowchart illustrating an operation of a verification unit provided in the character recognition device according to the second embodiment.

FIG. 5 is an explanatory diagram of a verification process according to the second embodiment.

FIG. 6 is an explanatory diagram of a character cutout method in a third embodiment of the character recognition device of the present invention.

FIG. 7 is a diagram illustrating another example of processing by the character cutout method according to the third embodiment.

FIG. 8 is a diagram illustrating an example of processing by a character cutout method according to the third embodiment.

FIG. 9 is a flowchart of processing of a character unrecognizable area in the fourth embodiment.

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 10 Character image 11 Character area extraction part 12 Feature extraction part 13 Dictionary collation part 14 Standard pattern dictionary 14_1 One character standard pattern dictionary 14_2 Multiple character standard pattern dictionary 15 Candidate character output part 16 Verification part 20, 20_1, 20_2 Pattern 21, 22 characters Code 30, 40, 50 Character image 60 Computer 61 CPU 62 Memory 63 Display device 64 Printer 65 CD-ROM drive 66 Magnetic disk 67 Mouse 68 Keyboard 69 Image scanner 70 System bus 100, 200 Character recognition device

 ────────────────────────────────────────────────── ─── Continuing on the front page (72) Inventor Shunji Sakane 4-1-1, Kamidadanaka, Nakahara-ku, Kawasaki-shi, Kanagawa Prefecture Inside Fujitsu Limited (72) Inventor Katsumi Ide 4-1-1, Kamiodanaka, Nakahara-ku, Kawasaki-shi, Kanagawa No. 1 Fujitsu Limited F-term (reference) 5B064 AA01 AB13 AB14 AB18 BA01 CA08 DA03 DA14 DC27 DC39 EA07 EA36

Claims (5)

[Claims] 1. A character region cutout unit that estimates a character width from an input character image and cuts out a character region having a length corresponding to the character width from the character image, and a character region cutout unit that cuts out the character region. A character extraction unit that extracts the characteristics of the character pattern in the character region, a one-character standard pattern dictionary in which the characteristics of each standard character are described, and the characteristics of each character string in which a predetermined number of standard characters are arranged A standard pattern dictionary composed of both a standard pattern dictionary including a plurality of characters, and a one-character standard pattern dictionary that constitutes the standard pattern dictionary by extracting the characteristics of the character patterns in the character area extracted by the characteristic extracting unit. And a character or a character string represented by a character pattern in the character area by comparing it with the characteristics of each character and the character string described in the plural-character standard pattern dictionary. A character recognition device, comprising: a dictionary matching unit that extracts a candidate of the character. 2. A method according to claim 1, wherein the dictionary matching unit extracts a character string candidate as a result of matching a characteristic of a character pattern in the character area with a characteristic of a character string described in the multi-character standard pattern dictionary. The feature extracting unit causes the character region in which the character string candidate is extracted to further extract the characteristic of the character pattern of each divided region obtained by dividing the character region into the same number as the number of characters constituting the character string. The character pattern feature of each of the divided regions is compared with the characteristic of each character described in the one-character standard pattern dictionary to extract a character string candidate corresponding to the arrangement of the divided regions, 2. The character recognition apparatus according to claim 1, further comprising a verification unit configured to verify whether the character string matches a character string candidate extracted using the plural-character standard pattern dictionary. 3. The character area cutout unit is configured to divide a character area having a length corresponding to the character width from the character image by a length equal to one division when the character width is divided into the same number as the predetermined number. 3. The character recognition device according to claim 1, wherein the character recognition device cuts out the characters while shifting them one by one. 4. The character string extracting unit extracts character or character string candidates for each of the character regions extracted by the character region extracting unit and integrates the extracted character or character string candidates. 4. The character recognition device according to claim 3, wherein the character recognition device extracts a candidate of the character. 5. The dictionary matching unit according to claim 1, wherein, when a character unrecognizable area in which character recognition is not possible remains, a characteristic of a character pattern in the character unrecognizable area is compared with a characteristic of each character described in the one-character standard pattern dictionary. The character recognition device according to claim 4, wherein the character candidate is extracted by collation.