JPS6126191A - Matching method of thick and thin pictures - Google Patents

Abstract

PURPOSE:To calculate the similarity or distance of thick and thin pictures having different photographing conditions each other by finding out the density grade vectors of respective picture elements of the thick and thin pictures to be compared and adding the difference of density grade vectors between the corresponding picture elements. CONSTITUTION:A picture input device 20 inputs the data of thick and thin pictures, a storage device 21 stores the data and an arithmetic unit 22 detects the density grade vectors of respective picture elements of the thick and thin pictures. The coordinates of a remarked pictures element are set up, the density data of the remarked picture element and its adjacent picture element are read from the storage device 21 to find out the difference between these data, the sum of the vectors is calculated to find out the density grade vector of the remarked picture element, and the storage device 23 stores the processed result of the arithmetic unit 22. A matching operation circuit 25 executes the matching operation between the density grade picture data of the inputted thick and thin pictures which are stored in the storage device 23 and the density grade picture data of a reference thick and thin pictures stored in the storage device 23. Namely, the distance or similarity between both the data can be calculated by adding the vector difference between the corresponding picture elements of both the data.

Description

[Detailed description of the invention] 〔Technical field〕 The present invention relates to a 1m light image matching method.

[Prior art]

The conventional grayscale image matching method uses the difference in density between grayscale images as a degree of similarity.

In the case of method 1:, even a simple shift in the overall density level causes a large sum of differences, making it impossible to detect the correct matching point. For example, although the grayscale image in Figure 2 and the grayscale image in Figure 3 have an overall density level shift of 2, it is clear that they are very similar (Figures 2 and 3). The numbers in the figure are concentration levels). However, even if these two grayscale images are superimposed, a small sum of density differences cannot be obtained.

As described above, in the conventional method, the photographing conditions for the two grayscale images must be the same or similar.

For the purpose of alleviating such constraints, a method has been proposed in which the maximum and minimum points of the density of a grayscale image are associated with each other, but none of these methods can achieve a sufficient effect. For example, the second
The grayscale image in the figure has one maximum point, whereas the grayscale image in FIG. 3, which is apparently similar to it, has two maximum points, resulting in a low similarity value between the two.

〔the purpose〕

An object of the present invention is to provide a method for matching gray scale images under different photographing conditions.

〔composition〕

The grayscale image matching method of the present invention calculates the density gradient vector of each pixel of two grayscale images to be compared, and adds the differences in the density gradient vectors between corresponding pixels. The feature is that it calculates degrees or distances.

That is, regarding the grayscale images in FIGS. 2 and 3.

In this method, density gradient images as shown in FIGS. 4 and 5 are obtained, and the two images are superimposed and matched. Note that the arrows in FIGS. 4 and 5 are unit vectors pointing in the direction of the maximum density gradient of the original image, and are shown here in eight directions as shown in FIG. 6(B). below,
The present invention will be described in detail.

FIG. 1 is a schematic block diagram showing one embodiment of the present invention. In this figure, 20 is an image input device for inputting grayscale image data, and 21 is a storage device for storing the grayscale image data.

22 is an arithmetic unit that detects the density gradient vector of each pixel of the grayscale image. FIG. 7 is a flowchart showing the process, FIG. 6 (A) is an explanatory diagram of the pixel of interest and its adjacent pixels, and FIG. 6 (B) is an explanatory diagram of the direction east vector.2 First, the pixel of interest (step 30), and determines whether the density gradient vector X of the pixel of interest X is undefined (step 31). If it has been defined, move the target pixel to the next coordinate. If it is undefined, the pixel of interest X and its adjacent pixel a+ b'+ C+ ((, e+
The density data of f+K+h is read from the storage device 21 and the following calculation is performed (step 32).

? - (ex) completion/1e-xl e= (c x) 2/ l c-x lb= (
b-x) 3/ l b-x l”'knee (a-x)
T/ l a-x lh= (h-x) 8/ I
h−x lHere, the denominator of each equation may be set to ■ or some other constant if the quality of the grayscale image is sufficiently good and stable.

By calculating the vector sum of a, b, . Then, te, -;, 7. −
Determine if all of H, Ma, Ryo, Ryo, and T are undefined 1
(Step 34) If undefined, it is determined whether processing of all pixels has been completed or not (Step 35); if not, processing proceeds to the next pixel.

Note that the method for determining the concentration gradient vector is not limited to the method described above, and may be changed as appropriate.

Returning to FIG. 1, 23 is a storage device for storing the processing results of the arithmetic unit 22. Reference numeral 24 denotes a storage device for storing density gradient image data of the reference gradation image. In this embodiment, the density gradient image data of the reference gradation image can also be written in the same path as the input gradation image to be compared with it. It has become. A matching calculation circuit 25 performs a matching calculation between the density gradient image data of the input gradation image stored in the storage device 23 and the density gradient image data of the reference gradation image stored in the storage device 24. That is, the distance (or similarity) between both pixels is calculated by adding the vector differences between the corresponding pixels. Specifically, the direction difference between the vectors is counted as a distance of m.

Note that one vector may not be stable.

For stabilization, it is recommended to average the concentration gradient vectors determined in the process shown in FIG. 7 within a small window. Especially when dealing with remote sensing images, it is desirable to average such density gradient vectors.

〔effect〕

As is clear from the above description, the grayscale image matching method of the present invention has the following effects.

(1) Since it is not easily affected by the overall shift of the m-degree bell, it is possible to match gray scale images taken under different photographing conditions.

(11) It is possible to perform more accurate matching than before even for grayscale images with uneven shapes, ings, and illumination.

(iii) It can also be effectively applied to matching three-dimensional grayscale images.

[Brief explanation of drawings]

FIG. 1 is a schematic block diagram showing one embodiment of the present invention, and FIG.
4 and 5 are schematic diagrams showing the density gradient images of the gradation images shown in FIGS. 2 and 3, respectively. Figure 6 (A) is a 3x3 pixel array diagram centered on the image of interest,
The same figure (B) is the direction! FIG. 7, which is an explanatory diagram of the It-order vector, is a flowchart of the process of determining the density gradient vector of each pixel. 20... Image input device, 21, 23.24...
-Storage device, 22...Arithmetic unit, 25...Matching arithmetic circuit. Figure 1 Figure 4 Figure 5 Figure 6 (A) CB>

Claims (1)

[Claims] (1) Find the density gradient vector of each pixel of the two grayscale images to be compared, and calculate the similarity or distance between the two grayscale images by adding the differences in the density gradient vectors between corresponding pixels. This is a grayscale image matching method that is characterized by: