JPH02214992A - Character recognizing device - Google Patents

Abstract

(57) [Summary] This bulletin contains application data before electronic filing, so abstract data is not recorded.

Description

DETAILED DESCRIPTION OF THE INVENTION Field of the Invention The present invention relates to a character recognition device that reads printed characters, dot characters, and handwritten characters from newspapers, magazines, etc., and converts input character patterns into code information such as JIS codes. It is something.

2. Description of the Related Art In recent years, character recognition devices have come to be used as simple character input devices. Conventional character recognition devices, for example, convert a rectangular character pattern to be recognized into MX
When dividing the character pattern into N meshes and scanning from the four sides of the character pattern to the opposite side, count the area of the background part until it encounters the character part, normalize this with the total area, and calculate this. (Umeda, Arino; “Classification of multi-font printed kanji using coarse peripheral patterns”, Institute of Electronics and Communication Engineers technical research report, PRL78-4)
.

Problems to be Solved by the Invention However, in the above-mentioned conventional technology, the shape of the peripheral part of the character pattern is expressed by the area of the background part.
This had the problem that the area of the background part also varied as the line width changed, and the area of the background part changed depending on the font and font.In other words, it was not a feature suitable for multiphone I compatible character recognition devices. . For example, Figure 4 shows examples of the character pattern ``Ki'' in Gothic typeface and Mincho typeface, respectively.
As shown in Figure 5, when these character patterns are divided into 4 x 4, the area of the background part is different as shown in Figure 8, and the area of the background part, that is, the feature amount is different from the font. were recognized as different characters from each other.

Means for Solving the Problems In order to solve the above problems, the present invention scans a recognition target character pattern cut out in a rectangular shape from an input image from the side of the rectangle toward the opposite side, and separates it from the background part. When a boundary pixel that changes to a character part is detected, and then a boundary pixel that changes from a character part to a background part is detected, the pixel located at the center of the country boundary pixel is found, and the pixel is detected in a plurality of predetermined directions at the pixel. The number of consecutive black pixels is counted, and the direction code of the direction in which the number of black pixels is the largest is recognized as a feature quantity.

Function: With the above-described configuration, the present invention expresses the characteristics of a character pattern using a directional coat, and identifies characters using this directional coat.

Embodiment FIG. 1 shows a functional block diagram of a character recognition device in an embodiment of the present invention. In FIG. 1, reference numeral 1 denotes an image input unit composed of a scanner or the like for reading image data;
Reference numeral 2 refers to a character cutting unit that cuts out the image read by the image input unit 1; 3 refers to a center pixel calculation unit that calculates the center coordinates of a character portion from the character pattern of each character cut out by the character cutting unit 2; and 4 A direction code extraction section detects a direction code based on the pixel obtained by the center pixel calculation section 3; 5 is a character recognition section that searches for a character coat corresponding to the direction code extracted by the direction code extraction section 4; 6 is a direction code extraction section; A dictionary 7 stores code combinations 7 and corresponding character codes in pairs, and 7 is a display section.

FIG. 2 is a block diagram showing the configuration of the character recognition device of this embodiment. Here, 21 outputs the character pattern read by a scanner that reads the character pattern as hit data. 22 is a RAM that stores bit data from the scanner 21, a character pattern area 23, a direction code area 24 obtained by analyzing the character pattern in this character pattern area 23, and a direction code stored in the direction code area 24. It has a character code area 25 to be determined.

Reference numeral 26 is a ROM and has a dictionary storing direction codes and character codes corresponding thereto, and a program storage area 28 storing a control program according to the flowcharts shown in FIGS. 3(a) and 3(b). 29 is a processing circuit that performs processing according to the control program stored in the program storage area 28. 30 is a keyboard for inputting data, and 6 is a display unit that displays characters corresponding to the character codes stored in the character code area 25.

The character recognition device of this embodiment configured as described above will be explained according to the flowchart shown in FIG.

First, in step S1, a character pattern to be recognized is divided into sub-character patterns. In step S2, 1 is substituted for i and the process moves to step S3.

In step S3, the character pattern area 23 is scanned downward from the pixel ((W/2N) X i , 0) of the character pattern. In step S4, it is determined whether or not there is a boundary point where the pixel value of the character pattern changes from O to 1, and if it exists, the scanning is further advanced and step S4 is determined.
5, it is next checked whether there is a boundary point that changes from 1 to O. If a boundary point exists in step S4 and step S5, the center pixel is calculated in step s6, and step S7
Proceed to. In step s7, the number of consecutive black pixels in the center pixel is counted in four directions, and in step s8, the direction code of the direction in which the number of black pixels is maximum is stored in the direction code area. If it is determined in step s4 that the boundary point does not exist in step s5, a code indicating that the boundary point does not exist is stored as a feature in step s15, and the process moves to step s9. In step s9, the pixel ((W
/2N) xi, scan upward from H). Then, in step slo and step sll, it is determined whether or not a boundary point exists, and in step s12, the center pixel is calculated, in step s13, the number of continuous black pixels is counted in four directions from the center pixel, and in step s14, the number of continuous black pixels is calculated. The direction code in the direction with the maximum number of pixels is stored in the direction code area 25 as a feature.

If the boundary point does not exist in step slO and step s11, a code indicating that the boundary point does not exist is stored in step s17. Then, in step s18, it is determined whether or not i is smaller than 4, and if it is smaller, step s1
At 6, i is incremented and the process returns to step S3, but if 1 is 4 or more, the process proceeds to step s19 and i is incremented.
Substitute. In step s20, the pixel (0, (W/2M
)xi) to the right. In step S21, it is determined whether a boundary point exists or not in step s22. If a boundary point exists, the center coordinates are determined in step s23, and the number of consecutive black pixels from the center pixel is counted in step s24. In step S25, the direction code of the direction in which the number of black pixels is maximum is stored in the direction code area 24 as a feature.

If no boundary point exists in step s21 and step s22, a code indicating that no changed pixel exists is stored in the direction code area 24 as a feature in step s33. Next, in step s26, the pixel (W, (W/
2M) Scan to the left from x i ), step s
27 and step 828, it is determined whether a boundary point exists or not, and step S30, step S31, or step s3
4, and in step s32 it is determined whether ] is smaller than 4. If it is not smaller, step s35
Then, add 1 to j and return to step s20, but if it is 4, the process ends.

The following operations will be explained using the recognition example character "tree" as an example.

First, the image input unit 1 inputs an image including a character pattern to be recognized as a binary image and stores it in the character pattern area 23 .

Next, the character cutting section 2 cuts out a character pattern to be recognized from the input image stored in the character pattern area 23. The pixel values of the character part and background part of the input character pattern are 1 and O, respectively. Furthermore, the pixels forming the character part and the background part are called black pixels and white pixels, respectively. A recognition example character pattern "tree" is shown in FIGS. 4 and 5.

Next, the center pixel calculation unit 3 calculates the center pixel of the character portion of the character pattern to be recognized. When the height and width of the rectangle circumscribing the character pattern are H and W, respectively, HXW is equally divided into M and %N blocks, and each block of the character pattern divided into MXN blocks is called a sub-character pattern. I'll decide. Here, the reason for dividing into equal parts is to simplify the explanation, and is not the essence of the present invention. Scanning is performed from the upper side of the rectangle circumscribing the character pattern toward the lower side, but at this time, scanning does not start from all pixels existing on the upper side (from pixel ((W/2N)xi, 0) Here, 1==1, 2. . . , N, and the divisor part is rounded off to the first decimal place and converted to an integer.By the above scanning, the boundary point where a white pixel changes to a black pixel, Next, a boundary point where a black pixel changes to a white pixel is detected, and a center pixel located at the center of the side boundary point is calculated.

Next, the direction coat extraction unit 4 scans the center pixel obtained by the center pixel calculation unit 3 in the four directions shown in FIG. The direction code in the direction with the maximum number of pixels is defined as the feature interval. If there is no boundary point, a code indicating this, for example 01, is set between features. Similarly, a total of 2x (M + N) features are extracted from the lower side to the upper side, from the left side to the right side, and from the right side to the left side. Since the character patterns in FIGS. 4 and 5 have H=W=24, the height and width of the Zabu character pattern are both 6. In this case, since there is no pixel in the center, the Zabu character pattern has the upper left as the origin (1, 1
>, when scanning from the top side to the bottom side of the rectangle circumscribing the character pattern, pixel (3,1), pixel (9,1), pixel (15,1), pixel 21°1) and scan downward until it hits the character part. In the case of pixel (3, 1,), the center pixel is (3, 7), and when calculating the number of consecutive black pixels in four directions at the center pixel, the number of black pixels with direction code 1 is 24. Since it is the largest, the direction coat with the largest number of black pixels, ie, 1, is set as the feature quantity. For the recognition example character pattern “tree” in Figures 4 and 5, M=4, N=
When divided into 4-parts, the data in the direction code storage area 24 as the extracted feature quantities of Gothic and Mincho fonts are completely equal, as shown in Fig. 7, and this method can be used for multi-font character recognition. It is valid.

Finally, the character recognition unit 5 compares the extracted feature values of the recognition pair ♀ character pattern with the feature values stored in the dictionary 6,
Candidate characters are determined and the results are displayed on the display section 7.

As described above, according to this embodiment, since the characteristics of the background portion of a character pattern are expressed by the direction coat, it is possible to accurately identify characters without depending on line width variations for each typeface.

Effects of the Invention The present invention expresses the characteristics of the background part of a character pattern using a direction code, so that regardless of the font of the character,
A character recognition device that can identify characters with high accuracy can be realized.

[Brief explanation of the drawing]

FIG. 1 is a functional block diagram of a character recognition device in one embodiment of the present invention, FIG. 2 is a block diagram showing the configuration of this embodiment, and FIGS. 3(a) and 3(b) are in this embodiment. FIG. 4 is a recognition example character pattern diagram for Gothic fonts, FIG. 5 is a recognition example character pattern diagram for Mincho fonts, FIG. 6 is a diagram showing the correspondence between direction and direction coat, and FIG.
The figure shows one part of the directional coat area extracted by this example.
FIG. 8 is a diagram showing the correspondence between the feature amounts of Mincho type characters and the feature amounts of Gothic type characters extracted by a conventional character recognition device. 1... Image input unit 2... Character cutting unit 3... Center pixel calculation unit 4... Direction code extraction unit 5... Character recognition unit 6... Dictionary 7... Display unit agent Name: Patent attorney Shigetaka Awano and one other person 1) Book

Claims (1)

[Scope of Claims] A character cutting section that cuts out a character pattern to be recognized into a rectangular shape from an input image; and a character pattern to be recognized when scanning from a side of the rectangle cut out by the character cutting section toward the opposite side. detection means for detecting pixels that change from a background part to a character part, and pixels on a character that become a boundary between a character part and a background part;
a center pixel calculation means for calculating a pixel located at the center of both boundary pixels; and a number of pixels on the character that exist consecutively in each of a plurality of predetermined directions from the center pixel calculated by the center pixel calculation means. a direction code extraction unit that counts the number of black pixels and extracts a direction code in the direction in which the number of black pixels is maximum, and collates the direction code extracted by the direction code extraction unit with the direction code stored in a dictionary. A character recognition device comprising: a character recognition unit that determines recognition candidate characters;