JPS6378287A - Character recognizing device - Google Patents

Abstract

PURPOSE:To discriminate small characters and ordinary characters by determining the result of recognition for objective characters of recognition by using the size ratio and positional relation of images of objective characters of recognition to surrounding images of the image of objective characters of recognition. CONSTITUTION:A character cutting-out section 2 cuts out an image of objective characters for recognition in a rectangular form from an image inputted on an image inputting section. A candidate of recognition extracting section 3 finds, for instance, the direction, number, position etc. of strokes of the image of objective characters of recognition obtained on the cutting out section 3 as the feature quantity of objective characters of recognition and collates with the feature quantity of each character registered beforehand on a dictionary, and extracts possible characters for recognition successively from nearer one. On the other hand, a character attribute determining section 4 determines the attribute of objective characters relation of image of objective characters of recognition to surrounding character images of objective characters of recognition obtained on the cutting out section 2. A result of recognition determining section 5 determines the result of recognition on the possible characters of recognition obtained on the extracting section 3 by using the attribute of objective characters of recognition determined by the determining section 4.

Description

DETAILED DESCRIPTION OF THE INVENTION (Field of Industrial Application) The present invention relates to a character recognition device that recognizes printed and handwritten characters from newspapers, magazines, etc., and converts them into pieces of information such as JIS codes, for example.

(Prior art) In conventional character recognition devices, character patterns are normalized by stretching and contracting the circumscribed rectangular area of a character part into a square of a certain size, and then feature quantities are obtained and recognition is performed (for example, , 1959 Institute of Electronics and Communication Engineers General National Association Tomoo Harada et al.
Handwritten kanji and hiragana recognition using weighted directional index histograms and pseudo-Bayesian identification m(II)J). In a device using this method.

For example, since both the input image shown in FIG. 5(a) and the input image shown in FIG. 5(b) are normalized as shown in FIG. 5(c), the recognition result depends on the size of the target character. It had the advantage of being less likely to be exposed.

(Problems to be Solved by the Invention) However, in the apparatus using the above method, for example, the image shown in FIG. 6(a) is normalized as shown in FIG. 6(b). In the case of small characters such as letters representing consonants and suffixes, punctuation marks, etc., normalization eliminates the difference in size from normal-sized characters, making it difficult to identify them.

In view of the above, an object of the present invention is to provide a character recognition device that can distinguish between small-sized characters and normal-sized characters.

(Means for Solving the Problems) In order to solve the above problems, the present invention uses the size ratio and positional relationship of the recognition target character image with respect to N character images surrounding the recognition target character image. The system is configured to determine recognition results for characters to be recognized.

(Function) With the above-described configuration, the present invention reflects information on the size and position of the characters in the recognition results, making it possible to distinguish between small characters and normal-sized characters. .

(Example) Examples of the present invention will be described below with reference to the drawings.

FIG. 1 is a block diagram showing an embodiment of a character recognition device of the present invention. In FIG. 1, "" is an image input section, into which an image containing characters to be recognized is input.

Reference numeral 2 denotes a character cutting section, which cuts out a rectangular character image to be recognized from the image input by the image input section 1. Reference numeral 3 denotes a recognition candidate extracting section, which extracts, for example, the recognition target character image that has been extracted by the character cutting section 2.

The direction of the stroke, the position of the number 1, etc. are determined as the characteristic quantities of the character to be identified, and are stored in the dictionary in advance in 9. The system compares the recorded feature values of each character and extracts recognition candidate characters in order of proximity. Reference numeral 4 denotes a character attribute determination unit, which determines the attributes of the recognition target character based on the size ratio and positional relationship of the recognition target character image with respect to character images surrounding the recognition target character image obtained by the character cutting unit 2. . 5 is a recognition result determination unit;
For the recognition candidate characters obtained by the recognition candidate extracting section 3, a recognition result is determined using the attributes of the recognition target character determined by the character attribute determining section 4. 6 is a display section.

Regarding the character recognition device of this embodiment configured as described above, which displays the recognition result obtained by the recognition result determination unit 5,
The operation will be explained below using an example where a horizontally written image is input.

The image input section 1 inputs an image as shown in FIG. 2(a). The character cutting section 2 divides the image input by the image input section 1 into character images to be recognized, each with a vertical width of 11.5 mm, as shown in FIG. 2(b). Cut out a rectangle with width w1. The recognition candidate extracting unit 3 checks whether or not each pixel of each recognition target character image is connected in each direction as shown in FIG. A direction code is assigned to the pixel of interest, and the connectivity of pixels is examined for each direction code to extract strokes. For example, in Figure 2 (
When the strokes of the recognition target character image C3 in b) are extracted, the result is as shown in FIG. 3(a). Then, find the direction of the stroke, the number 2 position, etc., and enter iZQ in the dictionary in advance.
The recognition candidate characters are extracted in descending order of distance by comparing them with the characteristic quantities of each character. FIG. 4 shows recognition candidate characters of the memorized recognition target character image (listed in order from left to right, starting with the first candidate). Character brim (in the gender determination unit 4, the vertical width 1. and the horizontal width W of the recognition target character image C) are the character images Ci-z r Ci-t + around the recognition target character image.
Average vertical width and average horizontal width W of Ci+1+C1o
Compared with l + < n X L and w, < mXW
(0<n<1°0<m<1), the position of the recognition target character image C1 is the surrounding character image cl-,,
c, -□rct*x+C If it is lower than the position of 1 or 2, it is a letter representing a consonant or a cursor, or a punctuation mark, or a letter written downward in small size such as ``''''.
The position of the recognition target character image C1 is a surrounding character image C1-2
If it is above the position of +C1-□+Ci or t+ci*z, the attribute of the character to be recognized is determined such that it is a character written in small letters upwards, such as ``''.

For example, if n=m=2/3, in the recognition target character image C1, 13<nX
L and 11 m x W, and the position of C1 is C, C,
, C4, and C3, the attribute of the character to be recognized is determined to be a character that is written in a smaller size downward. The recognition result determination unit 5 determines a recognition result for the recognition candidate characters extracted by the recognition candidate extraction unit 3 using the attributes of the recognition target character determined by the character attribute determination unit 4.

For example, for the recognition target character image C3, the recognition result is determined by setting "A" written in small size among the recognition candidate characters in the recognition target character image C1 shown in FIG. 4 as the first candidate.

In the explanation of the operation of the above example, the explanation was given for a horizontally written image, but when a vertically written image is input, the position of the character image to be recognized is on the left or right side of the character image surrounding the character image to be recognized. The recognition result can be similarly determined by determining whether

(Effects of the Invention) As explained above, according to the present invention, one punctuation mark, consonant,
Since it becomes possible to identify characters written in small size, such as characters representing cursive sounds, recognition accuracy can be improved.

[Brief explanation of the drawing]

Fig. 1 is a block diagram of a character recognition device according to an embodiment of the present invention, Fig. 2 is an explanatory diagram of an example of a method for determining recognition results in an embodiment of the present invention, and Fig. 3 is an explanation of a character recognition method. Figure 4 shows Lf! for characters to be recognized in one embodiment of the present invention! , illustrating candidate characters, and FIGS. 5 and 6 are explanatory diagrams of an example of a conventional method for normalizing a recognition target character image. 1... Image input section, 2... Character cutting section, 3.
... Recognition candidate extraction section, 4... Character attribute determination section, 5
... Recognition result determination section, 6... Display section. Patent applicant: Matsushita Electric Industrial Co., Ltd. Figure 1, Ward 3, Ward 4, Figure 5, Figure 6

Claims (1)

[Claims] an image input section for inputting an image including the recognition target character; a character cutting section for cutting out a recognition target character image in a rectangular shape from the image input by the image input section; and a recognition target character image obtained from the character cutting section. a recognition candidate extracting unit that calculates the feature amount of the recognition target character, compares the feature amount of the recognition target character with the feature amount of each character registered in advance in a dictionary, and extracts a recognition candidate character for the recognition target character; a character attribute determination unit that determines an attribute of the recognition target character based on a size ratio and positional relationship of the recognition target character image with respect to N character images surrounding the recognition target character image obtained by a cutting unit; A character characterized by having a recognition result determination unit that determines a recognition result using the attributes of the recognition target character determined by the character attribute determination unit for the recognition candidate character for the recognition target character extracted by the recognition candidate extraction unit. recognition device.