JPS63163685A - Direction code adding method - Google Patents

Abstract

PURPOSE:To save the time for direction code addition processing by applying raster scan by two picture elements each so as to reduce the number of times of reference of white picture elements of a white area of a character picture. CONSTITUTION:1st direction is used as the main scanning direction, the character picture is subject to raster scan by adjacent two picture elements each in the main scanning direction and when both the two picture elements are white or black, a specific direction code is added to both picture elements and when one of the two picture elements is the white level and the other is black, a direction code decided by adjacent four picture elements is added to the white picture element and a specific direction code is added to the black picture element, then a 2nd direction is used as the main scanning direction orthogonal to the 1st direction, the adjacent two picture elements are subject to raster scan with respect to the main scanning direction, and when one of the two picture elements is the white and the other is the black, and when the direction code added to the white picture element is a specific direction code, the processing to add correctively the other specific code is applied. Thus, the number of reference picture elements to decide the direction code is decreased and the direction code is added efficiently.

Description

DETAILED DESCRIPTION OF THE INVENTION [Technical Field] The present invention relates to a direction code adding method for adding direction codes to pixels of a character image.

[Prior art]

In a feature extraction unit of a character recognition device (OCR), a method is known in which a local direction code is attached to each pixel of a single character image in order to extract features from a binary image of a character.

The conventional direction code addition method is to raster scan a 5-character image pixel by pixel, and when the pixel of interest is white.

The direction code to be added to the pixel of interest is determined by referring to adjacent pixels in four directions (up, down, left and right) (or in eight surrounding directions), and because the number of pixels referenced to determine the direction code is large, processing speed is high. The problem was that it was slow.

The case where the direction code is determined by referring to the surrounding four pixels will be explained with reference to FIG. Coordinates of this figure (it j)
Consider the number of times the white pixel 11 is referenced.

Conventionally, in order to determine the direction code of a pixel of interest, a total of five pixels within the cross window 12 centered on the pixel of interest are referred to. Therefore, the coordinates (IIJI
) becomes the pixel of interest, the coordinate (x-1*j) becomes the pixel of interest as shown in (b) of the figure, and pixel 11 becomes the pixel of interest as shown in (C) of the figure. (d) in Figure 1
When the coordinates (1+l+j) become the pixel of interest as shown in .

Furthermore, as shown in (e) of the figure, the coordinates (l t j+ 1
) becomes the pixel of interest, each pixel 11 is referenced, and the number of times it is referenced is five times in total.

〔the purpose〕

Therefore, an object of the present invention is to provide a direction code addition method that can reduce the number of reference pixels for determining direction codes and efficiently add direction codes.

〔composition〕

The direction code adding method of the present invention raster scans a character image by two pixels adjacent in the main scanning direction with the first direction as the main scanning direction, and when both pixels are white or black, a specific direction code is added to both pixels. If one of the two pixels is white and the other black, a direction code determined by the four adjacent pixels is added to the white pixel, and the specific direction code is added to the black pixel. After performing raster scanning, two pixels adjacent to each other in the main scanning direction are raster scanned in a second direction perpendicular to the first direction as the main scanning direction,
When one of the two pixels is white and the other is black, if the direction code added to the white pixel is the specific direction code, processing is performed to re-attach another specific direction code. be.

〔Example〕

An embodiment of the present invention will be described below with reference to the drawings.

FIG. 1 is a binary image (11×11 pixels) of the character "+", and FIG. 2 is a diagram showing the correspondence between direction codes and patterns.

The direction code adding procedure in one embodiment of the present invention will be explained with reference to these two drawings.

First, from the upper left of the character image 13, in the i direction (horizontal direction)
The character image 13 is raster-scanned by two pixels arranged in the main scanning direction, with the character image 13 being set in the main scanning direction. That is, 2
A character image 13 is observed through a pixel window 14.

There are four types of two-pixel patterns within the window 14: ■white-white, ■white-black, ■black-white, and ■black-black.

In the case of patterns ① and ②, that is, when both pixels are white or black, direction code 0 is added to both pixels.

In the case of pattern (2), that is, when the left pixel of the two pixels is white and the right pixel is black, check the adjacent pixels above, below, and to the left of the white pixel on the left, and follow the direction code 1 according to the corresponding diagram in Figure 2. , 5.8 or 9 to that white pixel. That is, the direction code of this white pixel is determined by four pixels on the top, bottom, left and right, including the black pixel that is referred to at the same time.

A direction code O is added to the black pixel on the right side.

In the case of pattern ■, that is, the pixels on the left are black.

If the pixel on the right side is white, check the adjacent pixels above, below, and to the right of the white pixel on the right side, and set the direction code 3, according to the corresponding diagram in Figure 2.
Add 6, 7 or 9. That is, the direction code of this white pixel is determined by four pixels on the top, bottom, left and right, including the black pixel on the left side that is referenced at the same time. A direction code O is added to the left black pixel.

Through such direction code addition processing, a direction code as shown in FIG. 3 is added to each pixel of the character image 13. Note that, in this figure as well, the direction code of a pixel on which nothing is written is 0. As is clear from the figure, in this first stage of processing, direction codes 2 and 4 are not yet added.

Next, with the j direction (vertical direction) as the main scanning direction, raster scanning is performed from the upper left of the character image 13 through a window 15 of two pixels arranged in the main scanning direction.

The patterns within the window 15 are ■white-white, ■white-black.

There are four possible types: ■black-white, and ■black-black.

In the case of patterns ■ and ■, nothing is done to those two pixels.

In the case of pattern (2), the direction code added to the upper white pixel is checked, and if it is 0, direction code 4 is re-attached. If it is another direction code, just set it to N. Nothing is done to the lower black pixels.

In the case of pattern (2), the direction code added to the lower white pixel can be relied upon, and if it is 0, direction code 2 is re-attached. For any other direction code, do nothing. Nothing is done to the upper black pixels.

The result of such processing is as shown in FIG.

In this figure, the direction code of the pixel on which nothing is written is O.

2. This will be explained with reference to FIG. 5, which shows the number of times each white pixel in the white area is referenced. The white pixel 16 at coordinates (i, j) is detected when the window 14 is in the horizontal raster scan stage (a) and (b) of the figure.
Each is referenced once when the location is reached. Also,
At the stage of vertical raster scanning, the window 15 is referenced once each when it comes to the positions (Q) and (d) in the figure. In this way, since the two-pixel pattern that is simultaneously referenced in the white area is white-white, the number of times that the white pixel 16 is referenced is four times, which is one less than the conventional method.

Note that a white pixel (a white pixel in the outline of a character) whose two-pixel pattern that is simultaneously referenced is white-black or black-white is referenced five times as in the conventional case.

Now, in character image 13 in Figure 1, the total number of white pixels is 10.
8, and there are only 24 pixels in the outline of the character (pixels of the skewer mark). This tendency is true for character images in general.

Therefore, when the number of references to white pixels in the white area is reduced as described above, the number of references as a whole is significantly reduced, and the time for the direction code addition process is also significantly reduced.

Generally, a character image has approximately 64×64 pixels, but as the number of pixels increases, the effect of reducing processing time becomes greater.

FIG. 6 is a block diagram showing an example of an apparatus configuration for carrying out the above-mentioned direction code adding process, and FIG. 7 is a schematic flowchart showing an example of the processing operation.

In FIG. 6, 20 is a CPU, 21 is a main memory,
22 is an image memory, and 23 is a bus.

A character image 24 is stored in the image memory 22.

The main memory 21 stores programs and data for the direction code addition process, and the main memory 2 has a storage area 25 for storing direction codes added to character images.
1.

The processing operation will be explained below. Step 1 to Step 7
is a processing part that adds a direction code by horizontal raster scanning.

In step 1, initial values are set in i and j counters that specify the coordinates of pixels.

In step 2, data IMG (i
? j) and IMG (i-1, j) (white pixel = O
2 black pixels=1) are read out, subtraction is performed, and the result of the calculation is determined. If the calculation result is 0, that is, if the two-pixel pattern is the above or ■, step 3
Proceed to. When the calculation result is 1, that is, when the two-pixel pattern is the above-mentioned ■, the process proceeds to step 4. When the calculation result is -1, that is, when the two-pixel pattern is the above-mentioned ■, the process proceeds to step 5.

In step 3, the direction code of each pixel is stored in the storage area 25 as C0DE (i, j)=CODECx II
J)=O is written respectively.

In step 4, C0DE (i+, j)=O is written as the direction code of the right black pixel among the two pixels. Also, the data of the pixels to the left and above and below the white pixel are read out from the image memory, and depending on the pattern of the four pixels surrounding the white pixel, 1, 5.8 or 9 is the data C0DE.
It is written into the storage area 25 as (i-1,j).

In step 5, C0 is used as the direction code of the left black pixel.
DE (i-1,j)=O is written to storage area 25. Further, the data of the pixels to the right and above and below the white pixel on the right side are read out from the image memory 24, and 3°6.7 or 9 is the data C0DE depending on the pattern of the four pixels surrounding the white pixel.
It is written into the storage area 25 as (i, j).

Following steps 3, 4, or 5, a determination is made as to whether the horizontal raster scan is complete in step 6.

If it is determined that it is not finished, 1+J is determined in step 7.
Any necessary updates to the counters are made and processing continues from step 2.

If it is determined that the process has ended, the process proceeds from step 8 to step 15, which is a vertical raster scan process.

In step 8, initial values are set in the i and j counters.

In step 9, the data IMG is stored from the image memory 22.
(if j) and IMG (i, j-1) are read out and subtracted, and the calculation result is determined. 6 If the calculation result is O, that is, the 2-pixel pattern is the same as the above ■ or ■. If so, proceed to step 14. If the calculation result is 1, that is, if the two-pixel pattern is the above-mentioned ■', the process proceeds to step 10.

Further, if the calculation result is -1, that is, if the two-pixel pattern is the above-mentioned ■, the process proceeds to step 12.

In step 10, C0DE(II
j I) is read and its zero determination is performed. If it is zero, according to step 11, C0DE(i, j-1)
is rewritten to 4 and then proceeds to step 14.

In step 12, C0DE(II
j) is read and a zero determination is performed.

If it is zero, step 13 returns C0DE (i.

j) is rewritten to 2 and the process proceeds to step 14. In step 14, it is determined whether the vertical raster scan is complete. If it is determined that the i
, j counters are updated as necessary, and the process returns to step 9.

If it is determined that the process has ended, the direction code addition process ends,
The final processing result is obtained in the storage area 25.

Although one embodiment has been described above, the present invention is not limited to only the second embodiment, and can be implemented with various modifications.

(Effects) As is clear from the above description, according to the present invention, it is possible to reduce the number of references to white pixels in the white area of a single character image, and reduce the time required for direction code addition processing.

[Brief explanation of the drawing]

Figure 1 is a diagram showing a character image and its raster scanning method - Figure 2 is a diagram showing the correspondence between patterns and direction codes, Figure 3 is a diagram showing the results of direction code addition processing by horizontal raster scanning, and Figure 4 is A diagram showing the final processing result, FIG. 5 is an explanatory diagram of reference to white pixels in a white area, FIG. 6 is a block diagram showing an example of a device configuration for implementing direction code addition processing, and FIG. 7 8 is a schematic flowchart of the direction code addition processing operation, and FIG. 8 is an explanatory diagram of reference to a white pixel in a white area in the prior art. 13...Character image, 14...Horizontal raster scan window, 15...Vertical raster scan window. Figure 1 Figure 2 Figure 3 Figure 5 Figure 6 (Q) (b) (C) (d) (S)

Claims (1)

[Claims] (1) Raster scan the character image by two pixels adjacent in the main scanning direction with the first direction as the main scanning direction, and if both pixels are white or black, add a specific direction code to both pixels, If one of the two pixels is white and the other black, a direction code determined by the four adjacent pixels is added to the white pixel, and the specific direction code is added to the black pixel, and then the a second direction perpendicular to the first direction;
Raster scans two adjacent pixels in the main scanning direction with the direction of the main scanning direction as the main scanning direction, and if one of the two pixels is white and the other is black, if the direction code added to the white pixel is the specific direction code, For example, a direction code addition method characterized by performing a process of reattaching another specific direction code.