JPH0632079B2 - Character recognition device - Google Patents

Description

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a character recognition device for recognizing printed characters and handwritten characters of newspapers, magazines, etc. and converting them into character codes.

2. Description of the Related Art In a conventional character recognition device, in a recognition target character pattern cut out in a rectangle, each side of the rectangle used for the character cutout is roughly divided, and each side facing from each part of this divided side. The character background was scanned by scanning, and the histogram of the character background portion until the character portion was encountered was obtained, and the character recognition was performed based on this information. FIG. 2 shows recognition target characters “turn” and “enclosure” cut out in a rectangle. Each side of a rectangle obtained by cutting out these two characters is divided into four as shown in FIGS. Scanning each side facing each other from each of the four divided sides to obtain the histogram of the background portion until the character portion is encountered is as shown in FIGS. 3A and 3B. From the histogram of the background portion, Recognize.

Problems to be Solved by the Invention However, as is clear from FIGS. 3A and 3B, the histograms of the backgrounds corresponding to the respective sides of the cutout rectangle of the characters "time" and "enclosure" match, and the character "time" It turns out that it is difficult to distinguish between "and".

The present invention solves the above problems, and an object of the present invention is to provide a character recognition device that can identify characters having similar states around the characters.

Means for Solving the Problems In order to solve the above-mentioned problems, in the rectangle obtained by cutting out a recognition target character from an input image, each side facing from each side of the rectangle is scanned to obtain the recognition target. When a pixel point that changes from the background part of a character to a character part is detected, that pixel point is regarded as a change point, the pixel column number that is the number of the set to which the pixel point belongs is incremented by 1, and the pixel column number of that pixel point In the case of a pixel point other than the change point, a pixel column number one pixel before the pixel of interest is given, and a histogram of the number of pixels corresponding to the pixel column number is divided into a horizontal direction M and a vertical direction N for a total of M ×
It is obtained for each of the N sub-regions, and the value is recognized as the character feature amount.

Operation According to the present invention, by the above technical means, even for similar characters whose character background shapes are almost the same, the internal difference of the character is reflected in the histogram of the number of pixels corresponding to the pixel column number in the sub-region. It becomes possible to recognize.

Examples Hereinafter, examples of the present invention will be described with reference to the drawings.

FIG. 1 is a block diagram of an embodiment of a character recognition device according to the present invention. An image input unit 1 inputs an image including a recognition target character. Reference numeral 2 denotes a character cutout unit, which cuts out a recognition target character pattern in a rectangular shape from the image input by the image input unit 1. Reference numeral 3 denotes a region-specific density calculation unit that determines the pixel column number to which each pixel point belongs, and divides the rectangle cut out by the character cutout unit 2 into M in the horizontal direction and N in the vertical direction.
For each of the N sub-regions, a histogram of pixel points corresponding to the pixel column number is obtained. A recognition unit 4 extracts a recognition candidate character by comparing it with the dictionary 5 using the histogram obtained by the area density calculation unit as the feature amount of the recognition target character. 6
Is a display unit and represents the recognition result obtained by the recognition unit 4.

The operation of the character recognition apparatus configured as described above will be described by taking the recognition example character "" as an example.

The recognition example character "" is binarized by the image input unit 1. The character cutout unit 2 cuts out the rectangular pattern R shown in FIG. 4A as a recognition target character pattern from the binarized input image. The area-specific density calculation unit 3 scans each side of the rectangle R obtained by the character cutout unit 2 facing each other, sets the pixel row number of the pixel point on the side to 1, and recognizes the recognition target character pattern. When a change point which is a pixel point changing from the background portion to the character portion is detected, the pixel row number is incremented by 1 to obtain the pixel row number of the pixel point. In the case of a pixel point other than the change point, The pixel column number one pixel before is attached. FIG. 4 (b) shows the result of adding the pixel column numbers of all the pixel points when scanning is performed from the upper side of the recognition example character "". The cut-out rectangle is horizontally divided into four (M = 4) and vertically divided into four (N =
4) Obtain a histogram of each pixel point corresponding to the pixel column number for each of a total of M × N = 16 sub-regions. FIG. 5 is a histogram in the case of scanning from the upper side of the recognition example character “twice”, which is the characteristic amount of the recognition target character pattern. In the case of the recognition target character “turn”, when scanning is performed from the upper, lower, left, and right sides of the cut-out rectangle, a histogram similar to that in FIG. 5 can be obtained, and the histograms in the four directions of the upper, lower, left, and right are possessed as feature amounts.

The recognition unit 4 collates with the dictionary 5 which stores the feature amount of each character extracted by the same method as the recognition example character “twice”, extracts the recognition candidate character, and displays the result on the display unit 6.

In the present embodiment, when the recognition example character "enclosure" shown in FIG. 6 (a) is scanned from the upper side of the cut rectangle and pixel row numbers are added to all pixel points, as shown in FIG. 6 (b). Then, when the characteristic amount is obtained, the value becomes as shown in FIG. This value is different from the feature amount of the recognition example character "turn" shown in FIG. 5, and it is possible to recognize "turn" and "encircle" which are erroneously recognized in the conventional example.

In this embodiment, the histogram corresponding to all the pixel column numbers is obtained for all the rectangular sub-regions of the cut out recognition target character, but the histogram corresponding to the limited pixel column numbers is used for the limited sub-regions. However, it goes without saying that the same effect can be obtained.

According to the present invention, a recognition target character pattern cut out in a rectangle from an input image is divided into M in the horizontal direction and N in the vertical direction, M ×.
For each of the N areas, a histogram of the number of pixels corresponding to the pixel column number of each pixel point determined by scanning the opposite sides of the rectangle is obtained, and the value is recognized as a character feature. By doing so, it becomes possible to recognize even a similar character with a slight difference.

[Brief description of drawings]

FIG. 1 is a block diagram of a character recognition apparatus according to an embodiment of the present invention, FIG. 2 is a diagram showing recognition example characters "" and "", and FIG. 3 is a conventional method of "" and "". FIG. 4 is a diagram for explaining the feature amount according to FIG. 4, and FIG. 4 is an explanatory diagram of a method for obtaining the feature amount for each sub-region of “times” in the embodiment of the present invention.
FIG. 7 is an explanatory diagram of a histogram for each sub-region of “times” in the same embodiment, and FIG. 6 is a diagram showing another recognition example character “enclosure” in the same embodiment divided into sub-regions, and FIG.
The figure is an explanatory diagram of a histogram for each sub-region of the "circle". 1 ... Image input section, 2 ... Character cutout section, 3 ... Area density calculation section, 4 ... Recognition section, 5 ... Dictionary section, 6 ... Display section.

Claims (1)

[Claims] 1. An image input unit for inputting an image containing a recognition target character, a character cutting unit for cutting out a recognition target character pattern from the input image in a rectangle, and upper, lower, left and right sides of a cutting rectangle obtained by the character cutting unit. When the pixel point changing from the background part of the recognition target character to the character part is detected by scanning from each side facing each other, the pixel point is regarded as the change point, and the number of the set to which the pixel point belongs Is incremented by 1 to be the pixel column number of the pixel point, and in the case of pixel points other than the change point, the same pixel column number as the pixel point one pixel before the pixel point is given, A region-by-region density calculation unit that obtains each rectangle obtained by dividing the histogram of the pixel points corresponding to the pixel row number of each pixel point obtained by the character cutting unit into M in the horizontal direction and N in the vertical direction,
A character recognition device having a recognition unit for extracting candidate characters using the histogram obtained by the density calculation unit for each area.