JPH03228195A - Optical character recognizing device - Google Patents

Abstract

PURPOSE:To shorten the following recognition time of a type character by settling the classification of font in an original picture in accordance with the recognition category outputted from a recognizing part and information for collation. CONSTITUTION:A recognizing part 5 collates the recognition dictionary of a first font initially stored in a high-speed access storage part 6 with features of the character pattern extracted by a feature extracting part 4 and outputs the recognition category and information for collation to a recognition dictionary selecting part 7. A control part 9 repeats this processing with respect to one-row components of character patterns stored in a character pattern storage part 3 until collation with dictionaries of all fonts stored in a low-speed access storage part 8 is completed. When the processing is completed, a recognition dictionary selecting part 7 selects the recognition dictionary of an optimum font in accordance with the recognition category for each recognition dictionary and information for collation. The control part 9 stores the recognition dictionary of the reported font in the high-speed access storage part 6. Thus, the following collation time is shortened.

Description

DETAILED DESCRIPTION OF THE INVENTION [Field of Industrial Application] The present invention relates to an optical character recognition device (hereinafter referred to as OCR), and particularly to a multi-font printed character recognition device.

[Prior Art] Conventionally, OCR that recognizes printed characters in multiple fonts uses common features of each font category as a dictionary. One dictionary is prepared for one category, or multiple types of dictionaries are prepared for one category independently for each font.

[Problem to be solved by the invention]
In CR, according to the former method.

Since common features are used in the same category among multiple fonts, the reading accuracy is low and it is difficult to improve the recognition dictionary. Furthermore, the latter method has disadvantages in that it takes a long time to check the dictionaries of all fonts, and also requires a large memory to store the recognition dictionary, increasing the cost of the device.

The disadvantages of both of the above-mentioned methods become greater as the number of character types to be recognized increases. For example, if the recognition target is a Japanese document, the recognition target character types must be at least 3000 characters of JIS 1st level, and depending on the type of document, JIS
6,000 character types are also required at the IS second level, and the above-mentioned drawbacks make it difficult to implement as a device.

[Means for Solving the Problems] An object of the present invention is to focus on the fact that printed materials are usually printed in the same font, and to analyze all characters printed on the first page of printed materials for each font. The font type of the printed characters is determined based on the recognition result of the category at that time.
The object of the present invention is to provide an OCR that solves the problems of both of the above-mentioned methods by performing subsequent verification only with the dictionary of that font.

According to the present invention, there is provided an observation unit that photoelectrically converts an original image to obtain a character pattern for each character and outputs an input character pattern, and an observation unit that stores the input character pattern for one line and outputs the input character pattern for each character. a character pattern storage unit that extracts features of the stored character patterns and outputs the extracted features; a low-speed access storage unit that stores recognition dictionaries for a plurality of fonts; a high-speed access storage unit that stores, as a selected recognition dictionary, a recognition dictionary for one font selected by a selection signal from the recognition dictionaries for the plurality of fonts stored in an access storage unit; and the extracted feature and the selection. a recognition unit that outputs a recognition category and verification information by comparing the recognition dictionary with the recognition dictionary; and a recognition unit that determines a font of the printed matter to be recognized that corresponds to the stored character pattern from the recognition category and verification information, and determines the determined font. An optical character recognition device is obtained, which includes a recognition dictionary selection unit that outputs the font, and a control unit that outputs the selection signal from the determined font.

[Example] Reference will now be made to the drawings for embodiments of the invention.

This will be explained in more detail.

FIG. 1 is a block diagram showing the configuration of an optical character recognition device according to an embodiment of the present invention.

In FIG. 1, an observation section 2 outputs a character pattern for each character obtained by photoelectrically converting an original image 1 to a character pattern storage section 3. The character pattern storage section 3 is.

- A pattern memory that stores character patterns for lines, and outputs the character pattern to the feature extraction unit 4 character by character according to a control signal from the control unit 9. The feature extraction section 4 extracts the features of the character pattern sent from the character pattern storage section 3 and outputs them to the recognition section 5.

The recognition unit 5 compares the recognition dictionary of the first font initially stored in the high-speed access storage unit 6 with the characteristics of the character pattern extracted by the feature extraction unit 4, and uses the recognition category and information at the time of the comparison as a recognition dictionary. It is output to the selection section 7. The recognition dictionary selection unit 7 selects the recognition categories output from the recognition unit 5 and the information at the time of comparison.

- Store recognition categories for lines.

The control unit 9 performs control to repeatedly perform the above-mentioned processing on the character pattern for one line stored in the character pattern storage unit 3, and when the verification using the recognition dictionary initially stored in the high-speed access storage unit 6 is completed, Control is performed to store the second font recognition dictionary, which is previously stored in the low-speed access storage section 8, in the high-speed access storage section 6.

When the second font recognition dictionary is stored in the high-speed access storage unit 6, the control unit 9 repeats the same process as described above,
Check with the recognition dictionary of the second font.

Among the recognition categories and the information at the time of comparison, the recognition categories corresponding to - lines are stored in the recognition dictionary selection section 7.

These processes are repeated until all fonts stored in the low-speed access storage section 8 have been checked against the dictionary.

When the above processing is completed, the recognition dictionary selection unit 7 selects the most suitable font super-heavy dictionary based on the recognition categories from the first font recognition dictionary to the last font recognition dictionary and the information at the time of collation, and the control unit 9 Notify. The control unit 9 performs control to store the super heavy dictionary of fonts notified from the low-speed access storage unit 8 in the high-speed access storage unit 6 . At this point, the control unit 9 outputs the recognition category that has been checked with the dictionary of the most suitable font.

"Effects of the Invention" As explained above, according to the present invention, by determining the type of font in one original image based on the recognition category output from the recognition unit and the information at the time of matching, the subsequent recognition time of printed characters is reduced. This has the advantage that it is possible to reduce the size of the font, and the memory for storing the recognition dictionary can be configured with a memory capacity equivalent to one font.

[Brief explanation of drawings]

FIG. 1 is a block diagram showing the configuration of an optical character recognition device according to an embodiment of the present invention. 1... Original picture 12... View/I11 section, 3... Character pattern storage section, 4... Feature extraction section, 5... Recognition section,
6... High-speed access storage unit, 7... Recognition dictionary selection unit, 8... Low-speed access storage unit, 9... Control unit.

Claims (1)

[Claims] 1. An observation unit that photoelectrically converts the original image to obtain a character pattern for each character and outputs the input character pattern, and a character that stores the input character pattern for one line and outputs the stored character pattern for each character. a pattern storage section; a feature extraction section that extracts features of the stored character patterns and outputs the extracted features; a low-speed access storage section that stores recognition dictionaries for a plurality of fonts; a high-speed access storage unit that stores a recognition dictionary for one font selected by a selection signal from the stored recognition dictionaries for the plurality of fonts as a selected recognition dictionary; and the extracted feature and the selected recognition dictionary. a recognition unit that outputs a recognition category and verification information by comparing the recognition category and verification information; and a recognition unit that determines a font of a recognized printed matter corresponding to the stored character pattern from the recognition category and verification information and outputs the determined font. An optical character recognition device comprising: a dictionary selection section; and a control section outputting the selection signal from the determined font.