Fast image registration method based on parameters of image sub-blocks

Abstract

The invention provides a fast image registration method based on parameters of image sub-block. The method comprises the steps that firstly, a reference image is divided into the several image sub-blocks, and non-edge image sub-blocks with certain gray levels are selected from the image sub-blocks to serve as candidate reference image sub-blocks participating in registration; secondly, with the reference image sub-blocks as outer circulation, two inner circulation modules used for carrying out rough rotation angle matching and fine rotation angle matching respectively and a fine pixel position registration module are arranged in a nesting mode to complete image registration together, and the optimum parameters of registration image sub-block, namely, the rotation angle of a test image and the position parameters of registration test image sub-blocks, are provided; thirdly, the optimum parameters of the registration image sub-blocks are analyzed and utilized comprehensively to calculate and extract the maximum same image of the two images. By means of the method, the problem that a traditional image registration method based on the gray level is sensitive to factors such as gray level change of images, rotation of objects and sheltering of targets is greatly solved, and fast and accurate image registration is achieved.

Description

technical field [0001] The invention relates to the field of image processing, in particular to a fast image registration method based on image sub-block parameters. Background technique [0002] Image registration technology studies how to establish the mapping relationship between the source image (reference image) and the target image (test image) in space and gray scale. The essence is to find and establish the corresponding relationship between the physical position points of the same pixel in the two images. Image registration quality directly affects image content analysis, and is one of the core issues in the field of video content analysis. [0003] The image registration algorithm revolves around solving three problems: (1) quickly and accurately determine the possible relative displacement and rotation angle parameters between the two images; (2) reduce the gray level difference between the two images as much as possible The impact on image registration accuracy;...

AI Technical Summary

Problems solved by technology

[0007] The basic views on the image registration method based on feature points are: (1) have good adaptability to image grayscale changes, image deformation, local image occlusion, etc.; (2) matching information based on feature information The small amount can greatly reduce the calculation amount of the registration process; (3) The selection of feature points is a time-consuming and labor-intensive work, too few points, inaccurate points, or unevenly distributed points may be selected It leads to registration errors, and this situation often occurs; (4) Blurred images will make feature point extraction more difficult, and it is easy to miss feature points, resulting in false feature points, resulting in low registration accuracy

[0008] The traditional grayscale-based image registration methods mainly include partial pixel or block image registration. The main problem is the lack of an automatic selection of reference block images and a large-scale effective registration mechanism, because the selected reference block images Whether the location and the grayscale image information carried directly affects the accuracy and quality of image registration

For example, although the block image based on pixel sampling speeds up the registration speed, the image registration accuracy is not high because the original image information is not fully used.

Images