JPH01125683A - Character recognizing device - Google Patents

Abstract

PURPOSE:To recognize a character of any size with one datum, and to reduce a ROM memory quantity and a recognizing time by normalizing a character pattern and a character image to a constant size. CONSTITUTION:Character image information sent from a character extracting means is once stored in a RAM, thereafter, compared with character data in the ROM, and the character is recognized. A character recognizing means extracts a contour 8 of a character 7, and outputs the size of a circumscribed rectangle 9 with the use of the data of the outline 8. Since the proportion of enlargement and reduction at the time of normalization is obtained by the size, normalized character image data 6 can be obtained. The data are collated with normalized data 5 stored in the ROM. In the collation, the coincidence degree of the character is calculated, and the most coincident data out of the character data in the ROM are selected. Consequently, a single datum can recognize the character with any size.

Description

[Detailed description of the invention] [Industrial application field] The present invention relates to a character recognition device.

[Conventional technology]

There are many types of character recognition methods. Among them, the typical ones include pattern matching method and feature extraction method. Pattern matching method uses data with a certain size. The dot pattern is superimposed on the extracted character, the degree of matching with the dot pattern is checked, candidate characters are selected, and the most matching dot pattern is finally selected. Further, the feature extraction method is a method of picking out features such as points and glands in characters, and selecting candidate characters by hierarchically classifying them based on these features.

[Problem that the invention seeks to solve]

However, in the pattern matching method, although the algorithm is relatively simple, the types of characters that can be recognized with one butter/ are very limited, and there are many
Recognizing characters of type 1 requires a huge amount of data and, as a result, a large amount of memory. Another disadvantage is that the search time is very long. In addition, in the feature extraction method, the algorithm is very complex and the program is very long (and requires a lot of memory).

The present invention is intended to solve these problems, and its purpose is to reliably recognize many types of characters using a small amount of memory in a pattern matching method.

[Means for solving problems]

An input means for inputting a document image, an image processing means for processing the input image, a character string extraction means for extracting an input character string from the processed image, a character extraction means for extracting characters from the character string,
A character recognition device comprising a character recognition means for recognizing the characters and a recognized character display means is characterized in that the character recognition means uses character patterns and character images normalized to a constant size. Also, as a means of normalization,
This method is characterized in that a character circumscribing rectangle is extracted using the result of character outline extraction, the circumscribing rectangle is normalized to a constant size, and an internal character image is proportionally normalized.

〔Example〕

The present invention will be described in detail below based on examples.

Character recognition of the present invention! As shown in the block diagram of FIG. It is composed of 13 characters including character extraction means D for extracting characters, character recognition means E for recognizing extracted characters, and recognized character display means F for displaying recognition results.

FIG. 2 is a block diagram of the character recognition means E mentioned above.

The character image data sent from the character extraction means is once stored in the RAM 4, and then compared with the character data stored in the ROM 3 to recognize the characters.

The data stored in ROM 3 is a dot pattern of a fixed size.

The operation of the character recognition means E will be explained below based on the flowchart shown in FIG.

The character 7 extracted by the character extraction means is passed to the character recognition means E. Here, the outline 8 of the character 7 is extracted. A circumscribed rectangle 9 of the character 7 is extracted using the data of the outline 8. By knowing the size of the circumscribed rectangle 9, it is possible to know the appropriate size and reduction ratio during normalization, and normalized character image data 6 can be obtained. The character image data 6 is compared with the normalized data 5 stored in the ROi'vl 3. This comparison calculates the degree of matching of characters. In the next step, it is determined whether the matching degree is the highest among the collated character data so far, and if it is the highest, the corresponding character is rewritten and moved on to the next character data. If it is not the highest, it is compared with the next character data. When all the comparisons have been made, the character data stored in the ROM 3 with the highest degree of matching is selected. In the next step,
It is checked whether the degree of matching is greater than a certain value, and if it is less than a certain value, it is determined that it is not the relevant character, and if it is more than a certain value, it is determined that it is the relevant character, and the data is sent to the recognized character display means F. Character recognition is performed in this way.

As described above, according to the present invention, characters of all sizes can be recognized using one character data.

〔Effect of the invention〕

As described above, according to the present invention, during character recognition,
It is possible to make four faces of characters of any size with one character data. The character recognition method of the present invention is a pattern matching method for printed characters. In the pattern matching method, there is data corresponding to each character, and the degree of matching is checked and recognized by overlapping this data. Therefore, as the number of recognized characters and fonts increases, the amount of data inevitably increases. As a result, the amount of ROM memory increases and the recognition time increases. however,
By converting the size of characters to a constant size and comparing them as in the present invention, it becomes possible to recognize characters of any size using one piece of data. As a result, the effect of reducing the amount of memory in the ROM and the recognition time can be obtained.

[Brief explanation of the drawing]

FIG. 1 is a block diagram of a character recognition device of the present invention. FIG. 2 is a block diagram of the recognition means of the present invention. FIG. 3 is a diagram showing enlargement and reduction of a character image according to the present invention. FIG. 4 shows a flowchart of the recognition means of the present invention. A... Image input means B...vi image processing means C...Character string extraction means D...Character extraction means E...Character recognition means F... Recognized character display means 1...Character recognition means 2...CPU 3...ROM 4...RA ~1 5...Character data 6... Character image after normalization 7...Character image 8...Character outline 9...Character circumscribing rectangle that's all Applicant: Seiko Epson Corporation Agent: Patent attorney Mogami and 1 other person Figure 4

Claims (3)

[Claims] (1) An input means for inputting a document image, an image processing means for processing the input image, a character string extraction means for extracting a character string from the processed image, a character extraction means for extracting a character from the character string, and the character A character recognition device comprising a character recognition means for recognizing a character and a recognized character display means, wherein the character recognition means uses a character pattern and a character image normalized to a certain size. . (2) The scope of the claim characterized in that the normalization means is a method of extracting a character circumscribing rectangle, normalizing the circumscribing rectangle to a certain size, and proportionally normalizing the internal character image. The character recognition device according to item 1. (3) The character recognition device according to claim 2, characterized in that the result of character outline extraction is used as the circumscribed rectangle extraction means.