JPH01280887A - Character recognition method - Google Patents

Abstract

PURPOSE:To exactly recognize a plural characters-linking character by equipping the data of the plural characters-linking character in a dictionary and executing the recognition of the character by comparison with these dictionary data. CONSTITUTION:A hardware is composed of a picture input device 34 to input the document picture of a recognizing subject, a RAM 32 to accumulate picture information, which are inputted by the picture input device 34, and an arithmetic result, a CPU 31 to execute operation, a ROM 33 to store the dictionary data for the character recognition and the program of the operation and a display 35 to display a recognized result. Here, the data of the plural characters-linking character are equipped in the dictionary and the recognition of the character is executed by the comparison with these dictionary data. Thus, the plural characters-linking character, which is observed in an European and American document, etc., can be recognized without fail.

Description

[Detailed Description of the Invention] [Field of Industrial Application] The present invention is a character recognition device that inputs a document written on paper as an image, searches for a character area from the document image, and converts it into a code number. Regarding recognition methods.

[Conventional technology]

In recent years, rapid advances in character recognition devices have made it possible to automatically extract character areas from various document images, cut out and recognize individual characters, and automatically create document files. Various methods have been devised to recognize characters.

For example, as one of the character recognition methods, mesh features (Research and Practical Application Report, Vol. 34, No. 1, P, P.

47-57). This method expresses the rough needle-like distribution of the entire character. The feature extraction method involves dividing a circumscribed rectangle of a character to obtain nXn small regions. The area of the character part included in each of the small areas is counted and used as a mesh feature. The mesh feature is nxn per character.
The character can be estimated by comparing the ratio of the area of the character part in the area behind nXn with the data held as a dictionary.

In addition, as another method, peripheral characteristics (Research and Practical Application Report, Vol. 34, No. 1, p.

47-57). This method focuses on peripheral information of characters. The feature extraction method first finds a circumscribed rectangle of a character pattern, and divides each side of the circumscribed rectangle into n pieces. Next, it scans toward the characters from the divided side, and counts the area until it encounters the first character, and then the area until it encounters the next character. By performing the same processing on each divided side, it is possible to obtain peripheral features with nX4X2 data, and by comparing the nX4X2 data with the data owned by the dictionary and I2, the character can be specified. becomes possible.

Furthermore, most recognition methods, including the above two character recognition methods, cut out each character one by one before starting recognition. - To cut out each character, the vertical peripheral distribution of a character line is counted by 1, and the character break of each character is estimated and cut out from the information on the peripheral distribution and the size of the character.

[Problem to be solved by the invention]

The character recognition methods using peripheral features and mesh features have character data for each character, and recognition is performed by comparing this data with cut-out characters. It is extremely vulnerable to fluctuations in Particularly in recognition using the mesh feature, if another character is extracted together with the recognition target character, the position of the character within the circumscribed rectangle will shift, and there is a high possibility that the character will be recognized as a completely different character. For this reason, reliable cutting out of characters is required. However, when cutting out characters, there are many cases in which adjacent characters are connected, especially in European and American type printing characters, and separating these characters becomes a problem.

SUMMARY OF THE INVENTION The present invention has been made to solve the above-mentioned problems, and its purpose is to provide a method for reliably recognizing connected characters found in European and American documents.

[Means to solve the problem]

The character recognition method of the present invention reads a character image written on paper etc. using an optical image input means, recognizes the characters in the input image data, and replaces them with a code number. It is characterized in that characters are recognized by comparison with the dictionary data.

〔Example〕

Hereinafter, the present invention will be explained in detail based on examples.

The character recognition method of the present invention is generally used in hardware as shown in the block diagram of FIG. The hardware includes an image input device 34 for inputting a document image to be recognized, a RAM 32 for storing image information input by the image input device 34 and calculation results, and a CPU for executing calculations.
It is comprised of a PU 31, a ROM 33 containing dictionary data and calculation programs for character recognition, and a display 35 for displaying recognition results.

Hereinafter, the character recognition method of the present invention will be explained in detail according to the flowchart shown in FIG.

In block A, one character image to be recognized is input. At this time, the input character images vary depending on the time of reading by the image input device 34, but usually a plurality of lines of character images are input.

In block B, the peripheral distribution of the character image in the row direction is counted. By looking at the peripheral distribution, the position and line width of the character line can be known. Furthermore, the approximate size of the characters can be estimated from the line width.

In block C, characters are extracted from the character lines extracted in block B. Now, hypothetically, the character line extracted in block B is r 1denL shown in FIG. 21.
Assume that it is an image in which 14 characters such as 11'1catlonJ are written. When extracting each character from the characters 21, the peripheral distribution in the direction perpendicular to the character line as shown in FIG. 1 is usually counted.

The marginal distribution only indicates the presence or absence of characters, and is not an actual marginal distribution. By focusing on the peripheral distribution, the position and width of each - character can be found. Normally, when a plurality of characters are connected, the character width becomes abnormally large, so it can be determined that the characters are continuous and the characters can be forcibly cut off, making it possible to cut out a single - character. However, rldentINcatlo shown in FIG.
When characters are cut out according to the result of the marginal distribution 22, like the character image nJ, the character rfJ should be cut out at number 7 [1, but instead the character rfiJ is cut out. The character width of the character “fi” is the character rhJ as shown in Figure 2.
It has exactly the same width. Therefore, recognition proceeds without being able to determine from the character width that the character image rr iJ is two characters. When this character image is recognized, it will be recognized as h because it is very similar to the character rhJ. However, it is completely impossible to cut out the character rfiJ as two characters, and if you try to cut it out by force, there is a risk that the normal character rhJ will also be broken down. Therefore, in the present invention, recognition of block D is proceeded to next without forcibly separating such characters.

In block D, the characters cut out in block C are recognized. In the case of normal characters, recognition is performed without problems. However, it is impossible to recognize the character rfiJ shown in Figure 2, 21, and in most cases rhJ
I recognize that. Therefore, in the present invention, dictionary data used for recognition includes data of multiple connected characters as shown in FIGS. 1 to 5. Therefore, when character recognition is performed, it is possible to determine the character rfJ rtJ, and the second
In the image shown in Figure 21, each character is not completely cut out (but it is
It becomes possible to recognize it as atlonJ.

When character recognition is completed in the manner described above, the process of block C1 block D and further the process of block B1 block C1 block D is repeated until recognition of all characters in the input character image is completed.

When recognition of all characters is completed, the result is displayed on a display device such as the display 35 in block E, and the process ends.

〔Effect of the invention〕

As described above, according to the present invention, the dictionary used at the time of recognition has data for multiple character concatenation characters as shown in FIGS. 1 to 5 in the same form as data for - characters. . Therefore, originally, the characters shown in Figure 1 1 to 5 would fail to be cut out, and rhJ as shown in Figure 1 6.7,
It becomes possible to accurately distinguish characters that are mistaken for letters such as "m" from multi-character concatenated characters, and to give answers for a plurality of characters. Therefore, erroneous recognition due to failure in character extraction is greatly reduced, and highly sophisticated techniques are not required for character extraction, allowing high-speed character extraction. The present invention therefore greatly improves the reliability and speed of character recognition devices.

[Brief explanation of the drawing]

Figure 1 (a) and (b) are diagrams showing the dictionary of the character recognition method of the present invention, Figure 2 is a diagram showing the character extraction process, and Figure 3 is character recognition using the character recognition method of the present invention. A block diagram of the device and a flowchart of the character recognition method of the present invention are shown in FIG. 1 to 5...Multi-character concatenated characters 21...Extracted character line 22...Character line vertical peripheral distribution 31...Set number CPU 32/φΦ book/RAM 33...Dispute ROM 34...Image input device 35...Display A...φBlock A B...Block B C...Block C D Meeting...Block D E ...Block E Applicant Seiko Epson Co., Ltd. Agent Patent Attorney 1 t! Ill Miyabi Homare (Ike 1 person)
/ ? (d) (b) Figure 1 Figure 2 Figure 4

Claims (1)

[Claims] A character recognition method reads a character image written on paper etc. using an optical image input means, recognizes the characters in the input image data, and replaces them with a code number. 1. A character recognition method comprising data, and character recognition is performed by comparison with the dictionary data.