JP2503259B2 - How to determine full-width and half-width characters - Google Patents

Description

The present invention relates to a method for determining full-width and half-width characters in a character recognition device (OCR).

[Conventional technology]

Generally, the size of Chinese characters is called full-width, and the half width of full-width is called half-width. Character types that have both full-width and half-width include katakana, letters, and numbers. Therefore, full-width and half-width characters are recognized in the OCR by utilizing the fact that the conventional half-width characters are half the width of the full-width characters. That is, characters are cut out from the document image. In this process, the size of full-width characters is extracted. Then, the size (character width) of the cut out character is compared with the size of the full-width character to determine whether it is a full-width character or a half-width character.

[Problems to be Solved by the Invention]

However, in such a method, even if the full-width character is narrow, the katakana characters such as "i", "li", "to", the numeral "1", and the alphabetic character "I" are determined to be half-width. However, there is a problem that it becomes difficult to see when the full-width and half-width characters are mixed and output in the recognition result. For example, when there is a character row or column (character string) as shown in FIG. 2, the number "1" and the katakana "to", "ri", and "tsu" are recognized as half-width characters. In other words, existing OCR may incorrectly recognize double-width characters as narrow-width characters as single-byte characters.

Therefore, it is an object of the present invention to provide a method capable of preventing a narrow full-width character from being erroneously recognized as a half-width character.

[Means for solving the problem]

When a candidate character is extracted by performing unknown character recognition processing, it is determined whether the candidate character is a full-width character or a half-width character and its character type, and then a character having a half-width character for each character string. Check whether the character of the type exists, and if there is, extract the partial character string consisting of a series of characters of the same character type, check the number of full-width characters existing in the partial character string, According to the above, the corresponding partial character string is unified as a character string consisting of either one-byte characters or one-byte characters.

[Action]

For characters belonging to a character type in which full-width and half-width characters are mixed, not only the size (character width) of the character but also its context are checked to avoid misidentifying narrow full-width characters as half-width characters. When the reliability is improved, it guarantees a good appearance even when the output is "1".

〔Example〕

FIG. 1 is a flowchart showing an embodiment of the present invention.

First, the character image data is cut out by a known image processing (see), and in the process, the size of the reference double-byte character is extracted (see). Next, the size of the target character and the size of the full-width character are compared to determine whether it is a full-width character or a half-width character, and the result is stored in the memory as full-width half-width character information (see). Next, the target character is recognized (reference), and the result is converted into a predetermined character code and stored in the memory. In the process, the character type is determined by the character code. Then, by repeating the processes from to, one character row or column (character string)
Get the recognition result of.

Next, of the recognized character strings, for example, in order not to misidentify a full-width character with a narrow character as a half-width character, the characters belonging to the character type with half-width characters in the character string, that is, katakana, alphabet , The number is searched (reference), and if it exists, the characters starting from that character and before the character of another character type are extracted as one partial character string of the same character type (reference). At this time, the long syllabary of the character string starting with katakana (katakana string) is not the end of the character string. Similarly, in English strings, commas, periods, colons, and semicolons, and in numerical strings, neither commas nor periods end the string.

Next, check whether the full-width and half-width characters are mixed in this partial character string based on the full-width half-width character information obtained in step 2. If they are mixed, determine the number of full-width characters in the partial character string. Count (see). In this example, the number of double-byte characters is "0"
Only in case of, the corresponding character string shall consist of half-width characters, and in all other cases, the entire partial character string shall be unified as full-width characters (see).

Thus, for example, in the example of FIG. 2, even if the same result as the conventional one as shown in FIG. 2 is obtained by the processing of ~, the katakana string "trip" is extracted by the processing of, and the processing of In this case, since the double-byte character "p" is present in this partial character string, the entire partial character string "trip" is recognized (determined) as consisting of double-byte characters. As a result, "to", "ri"
It is possible to eliminate the possibility of erroneously discriminating a full-width character having a narrow width such as a half-width character, and it is possible to eliminate the discomfort when the recognition result is output.

〔The invention's effect〕

According to the present invention, for a character in which full-width and half-width characters are mixed, the full-width and half-width characters are determined by considering not only the size of the character (character width) but also its context. There is no risk of erroneously distinguishing a narrow full-width character from a half-width character. Therefore, if the reliability is improved, there is an advantage that it is possible to guarantee a good-looking image even if it is output.

[Brief description of drawings]

FIG. 1 is a flow chart showing an embodiment of the present invention,
The figure is an explanatory diagram for explaining an example of a character string and recognition results of full-width and half-width. Description of code: Character image cutout process, Full-width character size extraction process, Full-width / half-width determination process based on character size, Recognition result, Katakana, alphabetic and numeric search process, Katakana string, Extraction process of alphanumeric strings and numerical rates, ...
Counting process ... Unified processing.

Claims (1)

(57) [Claims] 1. A method of recognizing an unknown character to extract a candidate character thereof, and after determining whether the candidate character is a full-width character or a half-width character and a character type thereof based on the character width, For each character row or column (character string), it is checked whether or not there is a character type character having half-width characters, and if there is, a partial character string consisting of a series of characters of the same character type is extracted, The number of full-width characters existing in the partial character string is checked, and the partial character string is unified as a character string consisting of only one full-width character or one half-width character according to the number. How to determine.