JP2003162689A - Similar character recognition program and similar character recognition method - Google Patents

Abstract

(57) Abstract: The present invention relates to a similar character recognition program and a character recognition method for recognizing a character based on an image read from a character to be recognized. It is an object of the present invention to convert a candidate character based on the position information, the information of a rectangle constituting one character, etc., into a corresponding character when the candidate is a conversion candidate, to increase the determination conditions, and to improve the recognition rate of similar characters. SOLUTION: A means for generating a candidate by performing character recognition based on an image read from a character to be recognized, and a method in which the read character image to be recognized is below a reference line and a character j is included in the candidate Means for determining a character j when the character is included, and a character when the character image corresponding to one character of the read recognition target is separated into two or more and the character i is included in the candidate. A similar character recognition program and a similar character recognition method for causing a computer to function as means for determining i.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a similar character recognition program and a similar character recognition method for recognizing a character based on an image read from a character to be recognized.

[0002]

2. Description of the Related Art Conventionally, as shown in FIG. 5 (a), the English-U similar character identification is made up of whether or not a character rectangle is on a reference line (bottom tangent line of a character) and how many rectangles each character is. Similar characters are converted according to the characteristics such as whether or not they are used. For example, when the character rectangle is below the reference line (see (a-) in FIG. 5) when determining “i” and “j”, it is set to “j”, and “i” and “1” ( At the time of discrimination of (ell), if the rectangles are separated within the one-character rectangle as shown in (a-2) of FIG. 5, it is set as “i”, and the one-character rectangle is changed as shown in (a-3) of FIG. When not separated, a simple conversion such as “j” was performed.

Further, the attribute of the peripheral character of the target character is determined, and the target character is converted into a character having the same attribute as the peripheral character.
For example, when discriminating "1" (number is right) or "1" (ell), it is checked whether some characters before and after are alphabetic or numeric, and if it is a number, it is changed to "1" (ichi). If so, it was converted to "1" (L).

[0004]

Therefore, there is a problem that it is not possible to cope with the change in the character shape when the input image is a deteriorated image as shown in FIG. 5B.

In addition, conversion of similar characters based on information such as the position and number of one character rectangle gives a somewhat good conversion result, but there is a problem that many conversions are not practical.

Further, the above-mentioned conventional method has a problem that since the number of judgment conditions is small, conversion cannot be performed when one character rectangle has the same position and the same number. For example, there is a problem that all similar characters in Ei Uchi have an attribute of English, which is not suitable for identifying similar characters in Ei U.

The present invention solves these problems.
Regarding similar characters to be recognized, when they are candidates for conversion based on the position information of the reference line and the rectangle, the information of the rectangle that forms one character, etc., they are converted to the corresponding character, and the judgment conditions are increased to recognize similar characters. The purpose is to improve the rate.

[0008]

[Means for Solving the Problems] Means for solving the problems will be described with reference to FIG.

In FIG. 1, a character recognition system 1 recognizes a character based on an image read from a character to be recognized, and comprises a character recognition means 2 and a similar character recognition means 3. Is.

The character recognition means 2 searches the character recognition dictionary 4 for the image read from the character to be recognized and recognizes the character to generate a candidate.

The similar character recognizing means 3 determines a similar character based on the image corresponding to the read one character.

Next, the operation will be described. The character recognition unit 2 included in the character recognition system 1 searches the character recognition dictionary 4 for an image read from a character to be recognized, recognizes the character, and generates a candidate. The similar character recognition unit 3 reads the recognized character to be recognized. When the image is below the reference line and the character j is included in the candidates, it is judged as the character j, or the character image corresponding to one character of the read recognition target is separated into two or more. In addition, when the candidate includes the character i, the character i is determined.

At this time, when the image corresponding to one character of the read recognition target is not separated, but the character l (el) is included in the candidates, it is determined to be the character l (el). I am trying.

Therefore, the similar character to be recognized is converted into the corresponding character when it is a conversion candidate based on the position information of the reference line and the rectangle, the information of the rectangle forming one character, etc. By increasing the number, it is possible to improve the recognition rate of similar characters.

[0015]

BEST MODE FOR CARRYING OUT THE INVENTION Next, embodiments and operations of the present invention will be sequentially described in detail with reference to FIGS.

FIG. 1 shows a system configuration diagram of the present invention.
In FIG. 1, a character recognition system 1 executes various processes according to a program, and here, recognizes a character based on an image read from a character to be recognized. It is composed of the similar character recognition means 3 and the like.

The character recognition means 2 searches the character recognition dictionary 4 for the image of the character to be recognized, which is read from the document by the image input device 6, and recognizes the character to generate a candidate.

The similar character recognizing means 3 determines a similar character based on the image corresponding to the read one character.

The character recognition dictionary 4 is for registering character recognition information (character information corresponding to an image).

The output file 5 stores the recognition result. The image input device 6 reads an image of characters to be recognized printed on a document, and is a scanner or the like.

The output device 7 outputs the recognition result, and is a display device, a printing device, or the like.

Next, the configuration of FIG. 1 will be described in detail with reference to FIGS. 3 and 4 in the order of the flow chart of FIG.

FIG. 2 shows a flowchart for explaining the operation of the present invention. In FIG. 2, in step S1, characters are recognized after being read by the image input device. For example, a scanner, which is an image input device shown in FIG. 1, reads an image of a character to be recognized printed on a document, and the character recognition dictionary 4 is searched for the read image to perform character recognition. At this time, the reference line of the character image to be used in the subsequent processing is obtained and saved.
As a result, the image of the character to be recognized is read from the document, the character recognition dictionary 4 is searched based on the image, and the candidate character can be generated.

At S2, the recognition result is ・ I (eye) ・ 1 (L) ・ J ・ I The following processing is performed on the character that was.

In step S3, it is determined whether or not there is a character rectangle below the reference line. If YES, the process proceeds to S31. If NO, the process proceeds to S4.

In step S31, it is determined whether "j" is present in the candidates. If YES, it is determined to be "j" in S32. N
If O, go to S4.

In the process 1 consisting of S2, S3, S31 and S32 described above, when there is a character rectangle below the reference line and there is "j" in the candidates, it is judged as the character "j", It is possible to increase the recognition accuracy by increasing the number of recognition conditions (for details, refer to details of processing 1 in FIG. 3 described later).

Next, in S4, it is determined whether the recognition result is likely. In the case of YES, the character is determined to be probable. For example, when the correct reading rate is 80% or more as the character recognition result, it is determined that the character is recognized at this time. If NO, the process proceeds to S5.

In the process 2 consisting of the above S4, the recognition result is likely after the process 1 (correct reading rate is 80% or more, for example).
At this time, the recognition result can be determined (for details, refer to details of processing 2 in FIG. 3 described later).

In step S5, it is determined whether the one-character rectangle as the recognition result is separated into two or more. If YES, S51
Proceed to. If NO, the process proceeds to S6.

In step S51, it is further determined whether there is "i" in the candidates. If YES, it is determined to be "i" in S5. If NO, the process proceeds to S6.

In the process 3 consisting of S5, S51, and S52 described above, when one character rectangle of the recognition result is separated into two or more and there is "i" in the candidate, it is judged as the character "i". ,
It is possible to increase the recognition accuracy by increasing the number of recognition conditions (for details, refer to details of processing 3 in FIG. 3 described later).

In step S6, it is determined whether there is "f" or "t" in the candidates. If YES, it is determined to be "f" or "t" in S61. If NO, the process proceeds to S7.

In the process 4 consisting of S6 and S61 described above, it is possible to determine the "f" or "t" when the result is "f" or "t".

In step S7, it is determined whether the recognition result is "i" although the rectangle is not separated. If YES, S
Proceed to 71. If NO, the process proceeds to S8.

In step S71, "l" (L) is added to the candidates.
Determine if there is. If YES, in S72 "l"
Judge as (L). If NO, the process proceeds to S8.

In the process 5 consisting of S7, S71, and S72 described above, if one character rectangle is not separated but "l" (el) is present in the recognition result, it is determined to be "l" (el). Is possible.

In step S8, the result determined in the above processing 1 to 5 is output. By sequentially executing the above steps S1 to S7, the recognized candidates are further
By sequentially applying many conditions and determining a character that meets these conditions, it is possible to perform recognition determination under many conditions and improve the recognition accuracy.

3 and 4 are explanatory views of the present invention. Details of processing 1 to processing 5 in FIGS. 3 and 4 show specific examples of processing 1 to processing 5 in the flowchart of FIG. 2 described above.

-Details of processing 1: In the case of the character rectangle on the left side, and-Examples of YES in S31 of FIG. 2 are: -first candidate i-second candidate j, and judgment result "j It will be.

An example of NO in S31 of FIG. 2 is ・ First candidate i ・ Second candidate l And proceed to S4.

-Details of processing 2: In the case of the character rectangle on the left side, and-The example of YES in S4 of Fig. 2 is: -First candidate i correct reading probability 80% -Second candidate j correct reading probability 50 %, And the determination result is “i”.

An example of NO in S4 of FIG. ・ First candidate i 50% correct reading probability ・ Second candidate j correct reading probability 40% And proceed to S5.

-Details of processing 3: In the case of the character rectangle on the left side, and-The example of YES in S51 of Fig. 2 is: -First candidate j correct reading probability 70% -Second candidate i correct reading probability 50 %, And the determination result is “i” in S52.

An example of NO in S51 of FIG. 2 is ・ First candidate j, correct reading probability 70% ・ Second candidate l Probability of 50% correct reading Since there is no “i” in the candidate, the process proceeds to S6.

Process 4 details: -The example of YES in S6 of FIG. ・ First candidate i ・ Second candidate t And the determination result is "t" in S61.

The example 2 of YES in S6 of FIG. ・ First candidate i ・ Second candidate f And the determination result is “f” in S61.

An example of NO in S6 of FIG. ・ First candidate i ・ Second candidate l And there is no t or f, and the process proceeds to S7.

-Details of processing 5: In the case of the character rectangle on the left side, and-Examples of YES in S71 of Fig. 2 are: -First candidate i-Second candidate l (ell), determined in S72 The result is "l".

An example of NO in S31 of FIG. 2 is ・ First candidate i ・ Second candidate j And proceed to S8.

[0051]

As described above, according to the present invention,
As for the similar character to be recognized, a configuration is adopted in which it is converted to the corresponding character when it is a conversion candidate based on the position information of the reference line and the rectangle, the information of the rectangle forming one character, etc. It is possible to improve the recognition rate of similar characters by increasing the number of characters.

[Brief description of drawings]

FIG. 1 is a system configuration diagram of the present invention.

FIG. 2 is a flowchart explaining the operation of the present invention.

FIG. 3 is an explanatory diagram (1) of the present invention.

FIG. 4 is an explanatory view (No. 2) of the present invention.

FIG. 5 is an explanatory diagram of a conventional technique.

[Explanation of symbols]

1: Character recognition system 2: Character recognition means 3: Similar character determination means 4: Character recognition dictionary 5: Output file 6: Image input device 7: Output device

Claims (3)

[Claims] 1. A similar character recognition program for recognizing a character based on an image read from a character to be recognized, and means for recognizing the character based on the image read from the character to be recognized to generate candidates. A unit for determining the read recognition target character image as a character j when the character image is below the reference line and the character j is included in the candidates; The corresponding character image is separated into two or more, and the character i is included in the above candidate.
A similar character recognition program for causing a computer to function as a means for determining a character i when a character is included. 2. A character l (ell) is determined when the candidate includes a character l (el) even though the image corresponding to one character of the read recognition target is not separated. The similar character recognition program according to claim 1, further comprising: 3. A similar character recognition method for recognizing a character based on an image read from a character to be recognized, wherein the character recognition is performed based on the image read from the character to be recognized to generate candidates. Determining the character j when the read character image of the recognition target is below the reference line and the character j is included in the candidates; The corresponding character image is separated into two or more, and the character i is included in the above candidate.
The method for recognizing similar characters including the step of determining the character i when the character is included.