JPS60237583A - Correction system of character reading - Google Patents

Abstract

PURPOSE:To reduce unreadable characters by adding sequentially unreadable character patterns to a character recognition dictionary when unreadable characters are displayed and corrected through a display. CONSTITUTION:A character recognition means 26 reads a readout character pattern saved in a picture buffer memory 24 with the aid of a character recognition dictionary 25a, and in case of impossibility of reading said means 26 reads said pattern again with the aid of a rejected character recognitioi dictionary 25b. When said character pattern cannot be read in both readings, the unreadable character pattern is displayed on a display 3 by an editing processing means 27 through a user, and by keying in the character through a keyboard 4, said unreadable character is corrected. Simultaneously the editing processing means 27 registers the combination of the unreadable character pattern and the corrected character in the rejected character recognition dictionary 25b, whereby one character pattern for the recognition of the unreadable character is added.

Description

DETAILED DESCRIPTION OF THE INVENTION [Technical Field of the Invention] The present invention relates to a character reading correction method that can significantly reduce the number of unreadable characters in a single character reading device.

[Prior art]

As a conventional method of this scale, there is one described in Japanese Patent Laid-Open No. 56-21274 (FIG. 1).

In the figure, 1 is an OCR form, and 2 is a character reading device main body. 3 is a display device that displays the characters read by 2;
4 is a keyboard for correcting unreadable characters from the display 3;

Genomic phagocytosis means for reading the letters 1 in 21.

22 is a photoelectric conversion means, and 23 is a doji conversion means.

24 is an image buffer memory for storing read character patterns.

25 is a character recognition dictionary, and 26 is a character recognition means for recognizing character patterns stored in the image buffer memory 24 using this character recognition dictionary 25.

2γ displays the character pattern on the display 3.

This is an editing processing means for making corrections etc. using the keyboard 4. 28 is an output file for outputting character patterns.

Let me briefly explain the operation. The characters on the OCR form are optically scanned by the optical scanning means 21 and the one-source conversion means 2
2, it is converted into an electrical signal by quantization means 23, and a number of turns of read characters are created and stored in one image buffer memory 24.

Meanwhile, the character recognition means 26 interprets the read character pattern stored in the image reading memory 24 using the character recognition dictionary 25.

Here, 1 unreadable character turn is displayed on the display 3 by the editing processing means 27, the character is corrected using the keyboard 4, and the characters read by the character recognition means 26 are output to the output file 2B The sword is thrown out.

Since conventional character reading devices are configured to have 0 or more characters, they have the disadvantage that they can only read character patterns that are registered in advance in a single character recognition dictionary.

[Keystone of invention]

This invention was made in order to eliminate the above-mentioned drawbacks of the conventional system.1 When an illegible character is corrected from the display, the turn of the illegible character is added to the character recognition dictionary one by one. An object of the present invention is to provide a character reading correction method that can reduce the number of illegible characters.

[Embodiments of the invention]

An embodiment of the present invention will be described below with reference to the drawings. In the figure, the same reference numerals as in FIG. 1 indicate similar components. In FIG. 2, 25a is a character recognition dictionary prepared by the system, and 25b is a paste character recognition dictionary for storing illegible character patterns and corrected characters corrected by the editing processing means 27 for each user. It is.

The operation will be briefly explained. The characters on the OCR form are read by the optical scanning means 21, and the characters are read by the photoelectric conversion means 22. The operation in step 1 of storing the read character number turn in the image panzer memory 24 via the quantization means 23 is the same as in the prior art.

The device of the present invention first uses the character recognition dictionary 25a to decipher the read character number stored in the image buffer memory 24 by the character recognition means 26, and if it is unreadable, further reject character recognition is performed. It is interpreted using the dictionary 25b.

If the double-edged reading is still unreadable, the user can display the unreadable character pattern on the display 3 using the editing processing means 27, and then input the character using the keyboard 4. Make corrections.

At the same time, the editing processing means 27 registers the combination of the unreadable character pattern and the corrected character in the reject character recognition dictionary 25b, and adds one character pattern for read character recognition.

〔Effect of the invention〕

As described above, since this invention is configured to sequentially add otherwise illegible character patterns to the dictionary, the rejection rate of ordinary people who were repeatedly rejected in the first part has been greatly increased. This has the effect of providing a very agile character reading correction method that can reduce the number of characters and read with high efficiency.

[Brief explanation of the drawing]

FIG. 1 is a block diagram showing a conventional f-reading correction method. FIG. 2 shows the correction of character reading according to an embodiment of the present invention. It is a block diagram which shows a system. 1...OCR form, 2...Character reading device body, 3.
... Display device, 4... Keyboard, 21... Optical scanning means. 22... Photoelectric conversion means, 23... Quantization means, 24
...Image buffer memo'), 25a...Character recognition dictionary, 25b...Reject character recognition dictionary, 26...
Character recognition means. 27... Editing processing means, 28... Output file. In addition, in FIG. 1, the same reference numerals indicate the same or equivalent parts. Patent applicant: Mitsubishi Electric Corporation, -Tamaichi-11-----
−−−−−−−−−−～−Yo
----------Co ■ Written amendment to the musical proceedings (voluntary) Mr. Commissioner of the Japan Patent Office 1. Case description Japanese Patent Application No. 59-92911 2. Invention name character reading correction method 3. Person making the amendment 5. Amendment Column 6 of detailed description of the invention in the subject specification, contents of amendment

Claims (2)

[Claims] (1) In a character reading correction method in which an unidentifiable character image read from a recording medium is displayed on a display and corrected by operator intervention, A character reading correction method characterized by sequentially storing characters and using the unidentifiable character image as a character pattern for reading a recording medium. (2) A patent claim characterized in that the file storing the unidentifiable character image and the corrected characters of the unidentifiable character is initialized at the time of starting up the single character reading device so that each user can use the file. Character reading correction method described in scope 1.