JPS6324464A - Picture processing method - Google Patents

Abstract

PURPOSE:To extract a feature in a pattern recognition at high speed and with a high resolution by dividing a two dimensional original picture into plural blocks different in the number of picture elements and defining the average of the picture elements in the respective blocks to be the value of the block. CONSTITUTION:The two dimensional original picture constituted of the plural picture elements having the values is divided into the plural blocks different in the number of the included picture elements, the average value of the values of the respective picture elements included in the block is defined to be the value of the block in the respective divided blocks and the cluster of the values of the respective blocks is defined to be a new two dimensional picture. In the block having many picture elements, the high frequency component o the spatial frequency of the picture is interrupted and the value of a low frequency component is replaced by the representative of the block and since the block having the small number of the picture elements easily reproduce the original picture faithfully, a required picture area is represented by the block of the small number of the picture elements, thereby, faithfully reproduced and the picture of an unnecessary background part is represented by the block having many picture elements, thereby, the high frequency component is interrupted to compress the quantity of information.

Description

DETAILED DESCRIPTION OF THE INVENTION Field of the Invention The present invention relates to an image processing method for preprocessing a two-dimensional digital image for feature extraction necessary for pattern recognition and image understanding.

2. Description of the Related Art Recent image processing techniques tend to require high-speed and high-resolution processing.

Generally speaking, when dealing with the issue of high-speed, high-resolution image processing, the method is to compress the amount of information by passing the original image through a low-pass filter, finding the necessary local area, and until the desired resolution is obtained. The solution was to increase the resolution of that area.

An example of the conventional image processing method mentioned above will be described below with reference to the drawings.

4 and 5 show the multi-resolution structure of the conventional image processing method. In FIG. 4, 1 is an original image obtained by sampling a fixed field of view with nXn pixels. 2 is original image 1
For example, the average value of the 2×2 pixels above is set as one new pixel. This is a second layer image compressed to n/2 x n/2 pixels. Similarly, for 3 and 4, n/4\n/4, n
3rd layer image compressed to /8 x n/8 pixels, 4th layer image
This is a layered image. An image processing method for image data having a pyramid structure as described above will be explained using FIG. 6. In FIG. 5, 1 is the original image described in FIG. 4, and 2.degree.3 is the second layer image and third layer image.

4 is a target image for pattern recognition, and 5 and 6 are target images on the second layer image 2 and the third layer image 3, respectively. Extracting features from the top layer of the preprocessed image in this way reduces the time it takes to find a target image. This is because there is less image data to process. However, usually the resolution of the top layer may not satisfy the accuracy of the feature amount, so the required accuracy of the feature amount is determined by dividing the area from the position of the target image found in the top layer and moving to the third and second layers. The resolution is increased until the resolution is obtained, and finally, feature extraction is performed within a limited area of the original image 1.

Problems to be Solved by the Invention However, although the above-mentioned conventional technology certainly shortens processing time and achieves high-resolution processing, it is necessary to extract features for each layer, and when the target shape is large, had the problem that a large amount of pixel data had to be processed.

In view of the above-mentioned problems, the present invention aims to solve the problem that feature extraction must be performed using multiple hierarchical images created from an original image, which increases processing time and software size. It is something to do.

Means for Solving the Problems In order to solve the above problems, the image processing method of the present invention is as follows:
Divide a two-dimensional original image consisting of multiple pixels with values into multiple sections with different numbers of pixels, and calculate the average of the values of each pixel included in each divided section. The feature is that a value is taken as the value of that section, and a new two-dimensional image is created by having a set of values of each section.

Furthermore, the above-mentioned divisions are changed so that the number of pixels included in each division is made variable.

In a two-dimensional original image composed of multiple pixels with action values, in sections with a large number of pixels, the high-frequency components of the image's spatial frequency are cut off, and the values of the low-frequency components become representative of that section. Since it is easier to faithfully reproduce the original image using sections with a small number of pixels, necessary image areas can be faithfully reproduced by representing them with sections with a small number of pixels, and areas of the background that are not By representing an image as a section containing many pixels, the amount of information can be compressed by cutting off the high frequency components, thereby reducing the image processing time.

Furthermore, by making the sections variable, it is possible to arbitrarily change the faithful image reproduction location on the image.

Embodiment Hereinafter, an image processing method according to an embodiment of the present invention will be described with reference to the drawings.

FIG. 1 shows the partition distribution of an image divided into a plurality of partitions each containing a different number of pixels using the image processing method according to the first embodiment of the present invention. In FIG. 1, 12 is a 240 x 2400 pixel original image, and 13 is its constituent pixels. 14 is one of the sections in which the density average of the pixels in the section including 15×16 pixels is set as the new density of the area. As you move toward the center of the block, each side of the block decreases by one pixel, 14×14°13×13.

In this case, the number of density data is compressed to 1/120 of the original image. Also, the cutoff frequency near the center is high and reproduces the original image relatively faithfully, and the closer you go to the periphery, the lower the cutoff frequency is and the finer parts of the image are lost. Normally, pattern recognition targets are often processed near the center, and the amount of image data is increased near the center to reproduce finer parts of the image and improve the accuracy of feature values.

As described above, according to this embodiment, the low-pass filter is configured such that the number of pixels included in the original image decreases toward the center of the image, and the cutoff frequency increases toward the center. By doing so, it is possible to efficiently compress image data, and it is possible to achieve high-speed processing while maintaining resolution near the center and to reduce the hardware scale. In this embodiment, the number of pixels of the original image is 240 x 24 degrees, but the invention is not limited to this, and sections may be formed by reducing each side of the section by one pixel.

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

FIG. 2 shows a second embodiment of the present invention, in which the center position of the image processing method is changed when the sections are configured so that the number of pixels included in the section decreases toward the center of the image. FIG.

In the same figure, 16 is an image that has passed through a digital filter similar to the block distribution in which the two sides of the block are reduced by one pixel as in Figure 1, and 16 is an image that has been cut off by a block containing a small number of pixels. This is the point with the highest frequency and can represent even the thinnest part of the image, and 17 is the target image for feature extraction by pattern recognition. FIG. 3 shows the result of moving onto the target image a portion that can express even the thinnest portions of the image using sections containing fewer pixels.

The image processing method according to the second embodiment will be explained in order below.

As shown in Fig. 2, the target image 17 is a section that contains many pixels and can express only rough changes in the image, and is captured, and based on the position information, the image is A point 16 that can express the thinnest part is moved onto the target image 17. As a result, as shown in FIG. 3, an image is created in which the area around the target image faithfully reproduces the original image, and the closer to the periphery, the less detailed parts are expressed.

As described above, by moving a point on the target image that can express the thinnest part of the image with a section containing few pixels, it is possible to obtain information on the fine parts of the target image.
High-resolution feature extraction can be achieved for the target image.

Note that in this embodiment, the transfer function distribution is the same as in the first embodiment, but the number of pixels included in a section may be adaptively changed according to the shape of the target image.

Effects of the Invention As described above, the present invention divides a two-dimensional original image composed of a plurality of pixels having values into a plurality of sections each containing a different number of pixels, and calculates the average of the pixels in each section. By performing processing such that the area of the target image is represented as the value of the area, the area of the target image that is necessary is represented by a area that contains a small number of pixels to represent the target image as faithfully as possible, and the area around the target image that is not needed is By representing a portion as a section containing a large number of pixels, the image data as a whole can be compressed, and features such as pattern recognition can be extracted at high speed and with high resolution.

Further, by making this section variable, the part of the target image that is to be recognized as a faithful image can be made variable, and high-resolution image recognition can be performed as a whole.

[Brief explanation of the drawing]

FIG. 1 is a diagram showing the distribution of sections with different numbers of pixels included in the image processing method in the first embodiment of the present invention, and FIG. 2 is a diagram showing the distribution of pixels included in the sections in the image processing method in the second embodiment. Figure 3 shows the distribution of the blocks when the center is changed when the blocks are configured so that the number of pixels decreases toward the center. A diagram showing the result of moving the point that can express the thinnest part of the image onto the target image, Figure 4 is a diagram showing the multi-resolution structure of the conventional image processing method, and Figure 6 is an explanatory diagram of Figure 4. . 12...Original image, 13...Pixels, 14...
...Block, 16...Target image. Name of agent: Patent attorney Toshio Nakao and 1 other person No. 1
Figure (, 3 @ 2 Figure 3

Claims (4)

[Claims] (1) Divide a two-dimensional original image made up of multiple pixels with values into multiple sections each containing a different number of pixels, and then select the famous pixels included in the divided sections. An image processing method in which the average value of the values of is used as the value of that section, and a new two-dimensional image is created by having a set of values of each section. (2) The image processing method according to claim 1, wherein the divisions are configured such that the number of pixels included in the division decreases toward the center of the image. (3) Divide a two-dimensional original image made up of multiple pixels with values into multiple sections with different numbers of pixels, and in each divided section, each pixel included in that section The average value of the values of is set as the value of that section, and the set of values of each section is used to create a new two-dimensional image, and the section is changed from the original image to change the number of pixels included in each section. Image processing method made possible. (4) The image processing method according to claim 3, wherein the division is configured such that the number of pixels included in the division decreases toward the center of the image, and the center position is changed.