Stereo matching method based on translation and combined measurement of salient images

Abstract

The invention provides a stereo matching method based on translation and combined measurement of salient images. The method comprises that salient areas are respectively extracted from a left image and a right image shot by a binocular camera, and according to the salient images, a rough matching method based on translation of the salient images is provided to realize rough matching; and SAD (Sum of Absolute Differences) matching is implemented on the basis of rough matching, structural similarity (SSIM) is used for fine matching at the minimum of SAD, and the position with the highest SSIM score is selected as the best matching point. The stereo matching method can effectively improve the matching precision and reduce the computational complexity.

Description

technical field [0001] The invention belongs to the technical field of image processing, and in particular relates to a stereo matching method based on saliency map translation and joint measurement. Background technique [0002] Stereo vision is an important research field in computer vision. Its research purpose is to allow the computer to restore the three-dimensional scene through two or more two-dimensional images of the same scene and different angles, so as to recognize the three-dimensional world. Stereo matching is a key step in realizing stereo vision. Various stereo matching algorithms at home and abroad can be generally divided into three categories: feature-based stereo matching algorithms, global-based stereo matching algorithms, and local-based stereo matching algorithms. The feature-based stereo matching algorithm obtains a sparse disparity map, which requires interpolation estimation to obtain a dense disparity map. The energy function estimates the dispari...

AI Technical Summary

Problems solved by technology

[0004] At the same time, the methods disclosed in Document 1 and Document 2 are all based on local stereo matching algorithms. These methods are all for processing the entire image. In practical applications, only the information of the part of interest is generally needed, and the information of other parts can be ignored. Processing the entire image will increase a lot of unnecessary calculations, and the real-time performance is poor; the similarity discriminant function used in these methods is SAD, and the minimum value of SAD is the best matching point. However, when SAD is the minimum value As the best matching point, there will be the following problems: ①When the SAD reaches the minimum, it may not be the best matching point, or even a wrong matching point; ②The SAD at the best matching point may not be the minimum value; ③The minimum value of SAD has More than one, which SAD minimum value to choose as the best matching point is a problem

For this reason, taking the minimum value of SAD as the best point will generate a lot of wrong matching points, and the matching error rate is high, which is not conducive to improving the matching accuracy.

Images