JPS62115590A - Picture circumference length extracting system - Google Patents

Abstract

PURPOSE:To obtain a correct circumference length in a short time in a local and parallel type picture processing system by identifying the number for counting by the pattern of a local area consisting of picture elements of 3X3 centering the picture element. CONSTITUTION:A picture signal is inputted, stored and in a circumference length extraction of this picture, the number to be counted is obtained corresponding to the pattern consisting of the picture elements of 3X3, the numbers are tabulated, thereby the circumference length is obtained. For instance, in the respective picture elements of the picture, line i=1-8, column j=1-8 and the picture element is considered to be (ij). On the picture element 24, the local area is constituted by nine picture elements of 13, 14, 15, 23, 24, 25, 33, 34, 35. The picture element included in the picture is represented by one and that which is not included is represented by zero and a pattern code of a binary is obtained by a logical filter. Making this pattern code an address, the number to be counted is obtained from a converting table. This number is obtained on all the pictures, tabulated and the circumference length is obtained.

Description

[Detailed Description of the Invention] [Summary] The present invention relates to an image perimeter extraction method, and calculates the number of times to be counted when extracting the perimeter by 3x around the relevant pixel.
By identifying a local area pattern consisting of three pixels, the accurate perimeter can be extracted at high speed.

[Industrial application field]

The present invention relates to an image perimeter extraction method, and more particularly to a method for extracting an accurate perimeter at high speed in a locally parallel image processing method or the like.

The perimeter of an object or figure is extremely effective as a feature to be extracted for recognizing it. However, in local parallel processing systems, etc., extraction methods have been used that sacrifice accuracy in order to shorten processing time, and there is a demand for an extraction method that achieves both processing speed and accuracy.

[Conventional technology]

Although digital image processing has advantages such as being highly flexible and achieving high accuracy, it also has the disadvantage of requiring processing time because it processes each point sequentially. Various image processing methods are provided for this purpose.

The locally parallel image processing method can be said to be a method in which image processing functions are added to the image memory, and each pixel of the input image is sequentially scanned using the rask scanning method, and for each pixel, the local region including its neighborhood is For example, calculations for 3×3 pixels are processed in parallel.

Conventionally, when extracting the perimeter of the image shown in Fig. 4(a) using this method, the border line of this image is first extracted as shown in Fig. 4(b), and then the number of pixels of the border line is counted. In this example, the peripheral length is 18.

However, according to the definition of the perimeter, the perimeter is the value obtained by tracing the boundary line in the direction of the arrow, for example, and counting the number of pixels as shown in FIG. 5, and in this example, the correct value is perimeter = 20.

[Problem that the invention seeks to solve]

When extracting the perimeter of an image, the conventional method causes a large error as described above, leading to misrecognition, so it is necessary to develop an image perimeter that can obtain a value that matches the definition of the perimeter without increasing processing time. Extraction methods are strongly desired.

[Means for solving problems]

The problem is that there is a means for inputting an image signal, a means for storing the image signal, and a method for calculating the number of times to be counted when extracting the perimeter for each pixel of the image signal, with the pixel as the center.
This problem is solved by the image perimeter extraction method according to the present invention, which includes means for outputting a pattern corresponding to a local area pattern of ×3 pixels, and means for summing up the number of times to be counted.

[For production]

In the present invention, for an image signal that is input, stored, and whose boundaries are determined according to the prior art, 3×
Three pixels are defined as the local area, and the number of times to be counted when extracting the perimeter is identified based on the pattern of this local area, as illustrated in Figures 2(a) to (el). Indicate the included pixels with diagonal lines, (al) Example of a pattern that is not a border and is not counted:
This is a pattern in which all pixels in a local area are included in the image,
and various patterns in which the center pixel is not included in the image.

(b) Example of a pattern that should be counted once; (C) Example of a pattern that should be counted twice: (d) Example of pattern that should be counted three times: (e) Pattern that should be counted 4 times: It is.

That is, in contrast to the conventional method, which only divides pixels into two, into non-border pixels and border pixels, in the present invention, information on the number of times to be counted for border pixels is extracted at the same time.

If the added counting times are totaled, the pixels that should be counted twice or more are weighted and counted, and the correct perimeter according to the definition can be obtained.

Note that if the above processing is performed using a logical filter and a conversion table as shown in the embodiment, it can be performed with the same number of steps as the conventional process for determining a boundary line.

〔Example〕

The present invention will be specifically explained below using examples.

Each pixel of the image quoted earlier in Figure 4(a) is shown in Figure 1(
As shown in al.
．． It is composed of 9 pixels of 33.34.35.

Also, pixels included in the image are represented by 1", pixels not included in the image are represented by "0", and by a logical filter as shown in FIG. 2″, ..., the lower right pixel, for example, pixel 35, is multiplied by 2° and added, resulting in a binary number, in this example 10
0001011 represents the pattern of this local area.

The conversion table uses this pattern code as an address and the number of times to be counted when extracting the perimeter as conversion data. For example, this local area pattern 100001011 is a pattern in which the center pixel is counted once when extracting the perimeter; The conversion data at this address in the table is set to 1. For example, the local area pattern of pixel 44 is 10
1111001, which is a pattern in which the center pixel is counted twice during peripheral extraction, and the conversion data for this address in the conversion table is set to 2. Furthermore, for example, pixel 4
The local area pattern of No. 6 is 111111111, and the conversion data of this address is set to 0.

The converted data obtained in this way is as shown in FIG. 1 (bl), and if this is totaled, I X16+ 2 X 2 =20, and the correct circumference can be obtained.

〔Effect of the invention〕

As explained above, according to the present invention, it is possible to execute perimeter extraction using a locally parallel image processing method faithfully to its definition and without increasing the number of processing steps compared to the conventional method, thereby reducing the possibility of misrecognition, etc. Failures can be prevented.

[Brief explanation of drawings]

Fig. 1 is a diagram showing an embodiment of the present invention, Fig. 2 is a diagram showing an example of a pattern of a local area, Fig. 3 is a diagram showing a logical filter, and Fig. 4 is a perimeter determined by conventional boundary line pixel counting. Figure 5 is a diagram showing the definition of the perimeter. <a)' (b) Big example Nazu [Morning 1ku S rational fill L'7 or σ to brown 32 劃-numerical number f i no makoon/)) f'] 4 rows of non-crossing hessa pattern J 308 size)〈Sub σ pattern/) Chivalry J 40 battles 1 to ``S pa 2-ban castle is exploding I ash pattern 7 (guy doho 4 (2) Visit 52
to

Claims (1)

[Claims] A means for inputting an image signal, a means for storing the image signal, and a pattern of a local area consisting of 3×3 pixels with the pixel as the center, for each pixel of the image signal, the number of times to be counted when extracting the perimeter. An image perimeter extraction method characterized by comprising: means for outputting in response to the number of times to be counted; and means for summing up the number of times to be counted.