JPH07104908B2 - Image processing method - Google Patents

Description

Description: TECHNICAL FIELD The present invention relates to an image processing method for processing and coding an outline of image information.

[Prior Art] Conventionally, when the contour of image information is processed and coded, the contour line direction at each contour point is obtained as a quantized code according to the direction codes 1 to 8 shown in FIG. It was

However, in this method, when there is a pattern as shown in Fig. 2, for example, the direction code obtained when tracing the contour clockwise starting from S ₁ in the diagram is as shown in Fig. 3. , The direction code which originally changes continuously is A, in FIG.
As indicated by the portion B (corresponding to a and b in FIG. 2), the code becomes discontinuous, which causes inconvenience in recognition.

Further, as a means for preventing the above discontinuity, when the direction code changes discontinuously, such as when the direction code "8" changes to "1", as shown by () in FIG.
There is a method of maintaining continuity by adding / subtracting 8 to / from the direction code value. According to this method, in the case of the pattern shown in FIG.
The code is continuous as shown in FIG.

However, in this method, when the contour of the pattern is rotated once and returned to the start point of contour tracing, the code at that start point is the code obtained at the beginning as shown in S ₁ of FIG. It does not match.

In other words, the code values are discontinuous before and after the starting point. This means that even in the same pattern, the obtained code string changes depending on the position of the starting point. For example, if the starting point is S 2 in FIG. ₂ and the contour is traced clockwise, the direction code obtained when it is as shown in FIG. 4B is considerably different from that shown in FIG. It is different. This means that a change in the starting point due to the rotation of the original pattern or the like makes subsequent recognition difficult.

Moreover, finding discontinuities and adding or subtracting 8 to the code complicates such processing.

[Purpose] In view of the above points, the present invention eliminates the above-mentioned defects, captures the contours of image information as continuous data, and enables stable processing of contours of image information that is not affected by displacement or rotation of image information. An object is to provide an image processing method.

〔Example〕

Hereinafter, the present invention will be described in detail with reference to the drawings. The image referred to in the present application indicates characters, marks, symbols, pictures and the like.

FIG. 5 is a diagram showing an example of an image processing apparatus to which the present invention can be applied. It is not necessary to limit to such a structure.
Reference numeral 10 is a handy type image processing apparatus main body, 11 is a light source, and 12 is a document to be read. An optical system 13 guides light from the light source 11 and the pattern on the original 12 to a photoelectric conversion element (for example, a planar CCD) 14. Reference numeral 15 is a processing unit that processes and / or outputs output data from the photoelectric conversion element 14.

Next, the processing unit 15 will be described in detail. No. 6-1
The figure is a block diagram of the processing unit 15. A preprocessing circuit 16 performs binarization, noise removal, etc. on the input pattern.
Next, the characteristics of the pattern are extracted by the CPU 17 and the memory 18. The flow of feature extraction is shown in FIG. 6-2. That is, the contour information of the pattern is coded by the primary feature extraction. The code is used as it is depending on the pattern or the extraction method. Further, a feature different from the code string may be extracted by secondary feature extraction. Then, the feature obtained as described above is collated with a dictionary such as the memory 18 or the similarity or correlation is calculated to determine the pattern. Further, the above feature extraction may be repeated, or another feature extraction may be performed and then the pattern may be determined.

Next, the feature extraction of the present invention will be described in more detail.
FIG. 7 shows direction codes at each contour point to which the present invention is applied. In the case of the present embodiment, it is a code that represents the direction at each contour point when tracing the contour of the pattern clockwise. The tracking direction is not limited to this.

Here, the direction code at a certain contour point is determined by setting the contour line direction at the previous contour point to 0. That is, when tracing the contour of the pattern clockwise, the direction code at a certain contour point P _i is 1 when the contour line direction of P _i is 1
0 if it matches the contour line direction at the previous contour point P _i -1, and 45 °, 90 ° to the right as compared with the contour line direction of P _i -1
If it is rotated by 135 ° and 180 °, it is 1,2,3,4 and 45 to the left.
If they are rotated by ゜, 90 ゜ and 135 ゜, they are determined as -1, -2 and -3 respectively. For example, in the case of contour points P _{1 to} P ₁₁ as shown in FIG.
The contour line direction for each contour point is as shown by the arrow in the figure, and in this case the direction code is "0,0,1,0, −2,0,1,2,0".
Is decided. Here, the direction code for P ₁ cannot be determined unless the contour line direction of the preceding contour point is known, but generally when tracing the contour of the pattern, it always returns to the start point, so start at that time. The direction code of the point is always found.

When the direction code of the pattern shown in FIG. 2 is obtained by this method, it becomes as shown in FIG. As is clear from the figure, according to this method, the obtained code string is continuous. Furthermore, the code for the same contour point has the same value even after the contour has been circled. That is, it has periodicity with the number of contour points as the period. As a result, the obtained code string can be treated as a periodic function. In this case, even if the starting point of contour tracing changes, the phase of the periodic function is only shifted. For example, if S ₂ in FIG. ₂ is used as the starting point,
The code sequence starts from S _{2 in the} figure, but if we regard it as a periodic function, the function will only shift along the horizontal axis, and the shape of the function will remain unchanged.

Due to the above properties, the code string obtained by the present invention is
It is superior to those shown in FIGS. 3 and 4.

Next, FIG. 10 shows a flow chart for obtaining the code string described above. In step 1, the starting point of contour tracing is determined. Next, in step 2, the next first contour point is obtained. In step 3, the contour tangent direction is obtained from the start point and the first contour point. Next, in step 4, the m-th contour point is obtained (m = 2,3,4 ...), and in step 5, the contour tangent direction is obtained with the (m-1) th contour point. In step 6, the difference in the tangential direction of the contour is coded. Next, in step 7, it is judged whether or not the starting point has been returned to. If it has not returned to the starting point, the above operation is repeated.

By the way, the pattern can be identified by detecting, for example, the curved state of the contour (straight, right bend, left bend, sudden right bend, sudden left bend, ...) From the code string obtained here. it can. For example, the contour information corresponding to FIG. 11 can be obtained as shown in FIG. here,
Straight, right, left, and sharp right represent straight, right bend, left bend, and sharp right bend, respectively.

Furthermore, since the code string can be treated as a periodic function, it is possible to perform a discrete Fourier transform on the function and identify it on the frequency axis.

As described above, the method according to the present invention enables stable extraction of the contour information of a pattern, facilitates subsequent pattern identification processing, and conversely limits pattern recognition processing. not.

In the above description, the contour points for which the direction code is obtained are adjacent to each other. However, as shown in P _{1 to} P ₂₈ of the pattern of FIG. Alternatively, a method of obtaining a direction code may be used for contour points separated by a certain distance. According to this method, more stable contour features can be extracted without being affected by fine irregularities in the contour of the pattern.

In the above description, the direction code quantized in eight directions is used, but it may be subdivided as shown in FIG. 14, for example. The numbers in parentheses () shown in Fig. 13 are obtained according to the direction code shown in Fig. 14, and are able to more faithfully express the curved state of the contour, and a pattern set including similar categories. Suitable for pattern recognition in.

Needless to say, the methods described above may be repeated or combined.

The present invention may be applied to a system composed of a plurality of devices, may be applied to an apparatus composed of one device, or may be achieved by supplying a program to the system or the device. It goes without saying that it can be applied.

[Effect] As described in detail above, according to the present invention, the contour tracking start point of the image information when the contour of the image information is traced is determined, and the contour point with the image information and the previous contour point are determined. The contour tangent direction is obtained with the contour point, and the tangential direction is compared with the previously obtained tangential direction to derive and code a direction difference, and the derivation of the direction difference code is started by tracing the contour point. By repeatedly using the direction difference code derived when returning to the start point as the data of the start point, the contour feature of the image information can be stably expressed, and the position shift and rotation of the image information can be performed. It is possible to provide an image processing method that can easily obtain a continuous and periodic direction code string that is not affected by the above, and can reduce the contour processing of the image information.

[Brief description of drawings]

FIG. 1 is a diagram showing a conventional direction code, FIG. 2 is a diagram showing an example of a pattern, FIGS. 3 and 4 are diagrams showing a relationship between contour points and a direction code, and FIG. FIG. 6-1 is a block diagram of a processing unit, FIG. 6-2 is a diagram showing a flow of pattern feature extraction, and FIG. 7 is a diagram showing an example of an image processing apparatus to which the present invention can be applied. FIG. 8 is a diagram showing a direction code at each contour point to which the present invention is applied. FIG. 8 is a diagram showing an example of a pattern. FIG. 9 is a diagram showing a relationship between contour points and a direction code. FIG. 11 is a diagram showing pattern and contour information, FIG. 12 is diagram showing pattern code information and contour information, and FIG. 13 is a diagram showing an example of pattern. Figure 14 shows the subdivided direction code. 17 is a CPU and 15 is a processing unit.

Claims (1)

[Claims] 1. A contour tracing start point of image information is determined, a contour tangential direction is obtained from a contour point having the image information and a previous contour point, and the tangential direction is compared with the previously obtained tangential direction. Derivation of the obtained direction difference and coding, the derivation of the direction difference code is repeated until it returns to the starting point while tracing the contour points, and the direction difference code derived when returning to the starting point is used as the data of the starting point. An image processing method comprising: