Stereoscopic vision matching method based on weak edge and texture classification

Abstract

The invention discloses a stereoscopic vision matching method based on weak edge and texture classification, and solves a technical problem of achieving quick and precise compact matching. The method comprises the steps: carrying out the division and matching of edge and weak texture regions, carrying out parallax error correction and filling, carrying out the refinement of depth edge parallax errors and carrying out compacting of a parallax map. For MADC matching cost, the method carries out the matching according to different properties of regions, carries out mutual parallax error correction or filling according to a matching result, carries out parallax error refinement for a shade region and a parallax discontinuous region, and obtains the complete and compact parallax map. The beneficial effects of the invention are that the method does not need to carry out global optimization, reduces the operation burden of an algorithm, enables a parallax continuity constraint to be matched with the type of the region, and avoids a conflict of common local optimization with the continuity constraint in the prior art. The method is high in universality, and can be widely used for various types of binocular vision and multi-view 3D imaging algorithms.

Description

technical field [0001] The invention relates to the fields of image processing and machine vision, in particular to a stereo vision matching method based on weak edge and texture classification. Background technique [0002] In the process of binocular stereo vision measurement, after obtaining the corresponding relationship of image points through stereo matching, the corresponding disparity map can be obtained, combined with the projection imaging model determined in the calibration process, the depth information of the corresponding point can be determined through calculation. Due to the sparsity and discontinuity of the feature itself, the feature-based stereo matching algorithm cannot directly obtain the corresponding relationship of non-feature pixels, but can only obtain a sparse disparity map, which requires a subsequent complex interpolation process to generate a dense disparity map. If you want to directly obtain a dense disparity map, you need to select a matching...

AI Technical Summary

Problems solved by technology

[0013] The global algorithm with high matching accuracy often has a very large amount of calculation, which makes it difficult to meet the real-time performance of the algorithm; while the local algorithm with good real-time performance lacks effective optimization for disparity continuity, the matching accuracy is often low; Improving computational efficiency while maintaining accuracy is an important issue faced by stereo matching algorithms.

Images