JPH0368091A - Character recognizing device - Google Patents

Abstract

PURPOSE:To shorten the time required for character recognition by storing data of the number of all test bits in a character pattern data dictionary. CONSTITUTION:Pairs of character patterns 401, 403, and 405 and numbers 402, 404, and 406 of total comparison picture elements are written as follows; the character pattern data dictionary is written in the part 401 and the number of total comparison picture elements in the character pattern data dictionary is written in the part 402 and the character pattern data dictionary of the next character is written in the part 403 and the number of total comparison picture elements of the character is written in the part 404. Consequently, only numbers of the coinciding picture elements between character pattern data 401, 403, and 405 and an extracted character are counted, and numbers 402, 404, and 406 of total comparison picture elements are read in at the time of the end of comparison of all picture elements and are counted to obtain degrees of coincidence. As the result, the counting instruction executed for comparison of each picture element is omitted.

Description

[Detailed description of the invention]

[Industrial Field of Application] The present invention relates to a character recognition device that extracts a character area from a single document image by manually combining the images of characters written on a sheet of paper and converting the extracted character area into a code number.

[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, the pattern matching method is the most basic method for character recognition. In this method, as a dictionary for character recognition, character patterns of the remaining target characters are stored in the dictionary, and at the time of recognition, by superimposing the extracted character pattern and the character pattern in the dictionary, The degree of matching is checked and characters that match are recognized characters. Although this method is very simple, if the fonts of the two recognition target characters match the characters in the dictionary, they are recognized accurately at a high rate. Also, the Aii notation method is usually 2
Although the extracted characters are compared as they are, they may be normalized and used to absorb the impact caused by the size of a single character. Furthermore, in order to eliminate discrepancies caused by differences in fonts, character patterns are given a certain degree of freedom. [Problem to be Solved by the Invention 1] However, although the above method is a very simple method and has the advantage of fairly accurate recognition, it is difficult to distinguish between the extracted character pattern and the character pattern in the dictionary. Because there are so many pixels to compare, recognition takes a long time. Character recognition devices have serious challenges. SUMMARY OF THE INVENTION In order to solve the above-mentioned problems, the present invention aims to provide a character recognition device that recognizes characters as little as 6 times faster in a character recognition method using pattern matching. [Means for Solving the Problems] The character recognition device of the present invention includes an optical image input means for inputting a document image by photoelectrically converting light reflected from a sheet of paper, and positions of character lines and words from the input image. means for detecting and extracting each word, and a character recognition means for extracting each character from the extracted characters, comparing it with pre-existing character pattern data, and converting it into the highest numerical character code. A character recognition device having the following is characterized in that data of the total number of test bits is recorded in a single character pattern data dictionary. [Example 1] The present invention will be described in detail below based on an example. A block diagram of the character recognition device of the present invention is shown in FIG. 1. The character recognition device executes processing according to a program.
puiot, image input device 1025 for inputting character images into a storage device; character display means 103 for displaying character recognition results;
ROM104 containing a 5-character pattern data dictionary
, and RAM l, which is a storage device that stores character images.
It is composed of 05. The character recognition method of the character recognition device of the present invention will be explained below based on the flowchart shown in FIG. 6.
8 The character recognition device of the present invention, which will be explained in detail with reference to the drawings and FIG. To extract a word region from the zero-order input character image input to the RAM 105, the peripheral distribution in the row direction of the input character image is first counted. The value of the marginal distribution (not shown) increases at the position where character lines exist, and the value of the marginal distribution is small between character lines. Therefore, the position of the character line can be determined by the value of the marginal distribution. After estimating the position of the 0 character line, which can be easily estimated, the peripheral distribution in the direction perpendicular to the line direction of the estimated character line is then counted. Areas where the value of the marginal distribution (not shown) is large are areas where characters exist, and areas where the value is small are areas where no characters exist. Therefore, by examining areas where no characters exist, the sizes of word spacing and character spacing can be estimated, and word areas can be extracted. Once the word area has been extracted, the next step is to recognize the extracted word.
Character recognition by conventional pattern matching, which is performed by comparison with the character pattern data dictionary stored in Candidate characters are narrowed down by checking the degree of matching. In addition, as shown in Figure 2, the format of the dictionary is such that when the extracted characters are normalized and compared, the part 203 that always becomes the character part of any character can absorb slight positional deviations and differences in font. ,
A dictionary formed of three areas: a part 201 that is always a blank part for any character, and a part 202 that is not subject to comparison because it becomes a character part or a blank part depending on the character, or simply, The third one expresses the pattern of the target font as it is so that only the target font is recognized.
The format is as shown in Figures (a) and 3(b), and the number of pixels compared for each character is different. Therefore, when recognizing characters using the character pattern dictionary, the character pattern While comparing pixel by pixel of the data dictionary and extracted character data, count the total number of matching pixels, at the same time count the total number of comparison pixels, and finally count the degree of character matching and narrow down to one character. Therefore, in the present invention, a character pattern data dictionary as shown in FIGS. 4 and 5 is provided. Figure 4 shows the part 4 in Figure 2 where the character Bakun data dictionary is written.
01, and a portion 402 in which the total number of comparison pixels in the character pattern dictionary is written. Further, one character pattern data and the total number of comparison pixels are written as a set, such as a character pattern data dictionary 403 of the next character and a total comparison pixel number 404 of the character. Similarly to FIG. 5, the character pattern data dictionary 5 shown in FIG.
01,503.505 and total number of comparison pixels 502
．． 504.506. Therefore, when comparing with single character pattern data, count only the number of pixels that match the single character pattern data and the extracted character, and when all pixels have been compared, read and count the total number of comparison pixels. The degree of agreement is determined. As a result, the counting instruction that is performed every time each pixel is compared can be omitted. In pattern matching, all pixels in the extracted character pattern and the dictionary are compared, and since each character has a large number of pixels (usually more than 1000 pixels), the reduction in the number of instructions can save a considerable amount of time. can be reduced. Furthermore, when recognizing words, many characters are compared against several characters, so the difference in speed becomes large enough to be felt by humans. Therefore, by using the recognition method of the present invention, it is possible to greatly shorten the time required for character recognition in the pattern matching method. The words recognized as described above are displayed on the recognition result display device 103.
The recognition operation is completed by outputting the output at . [Effects of the Invention] As described above, in the character recognition method using pattern matching, the present invention has the total number of comparison pixels in the character pattern data dictionary, so the total number of comparison pixels when comparing pixel by pixel is There is no need to count the number of comparison pixels. As a result, the time required for recognition can be significantly reduced, and the performance of the character recognition device can be significantly improved.

[Brief explanation of drawings]

FIG. 1 shows a block diagram of a character recognition device according to the present invention. FIGS. 2, 3, 4, and 5 are diagrams showing an overview of the present invention. FIG. 6 shows a flowchart of the character recognition device of the present invention. 101...CPU 102...Image input device 103...Recognition result display device...ROM...RAM 403, 40 402, 404. 40 503, 50 504, 50 501°502, 04 05 01 02 03 01 02 401.・Common white area ・Non-comparison target area ・Common character area ・Text area ・White space area ・Total comparison pixel countTotal comparison pixel countCharacter pattern dataCharacter pattern data or more

Claims (1)

[Claims] an optical image input means for inputting a document image by photoelectrically converting light reflected from a paper surface, a means for detecting the positions of character lines and words from the input image and extracting each word, and a means for extracting each word from the extracted characters. In a character recognition device equipped with a character recognition means that extracts each character, compares it with pre-existing character pattern data, and converts it to the character code with the highest degree of matching, all tests are stored in the character pattern data dictionary. A character recognition device characterized by recording bit number data.