An Improved Semi-Global Stereo Matching Algorithm Based on Multi-cost Fusion
Wenlong Li, Ranran Hu, Miao Gao · 2021
As the matching accuracy of the traditional Semi Global Matching (SGM) algorithm is not satisfactory in illumination and weak texture areas, a SGM algorithm based on multi-cost fusion is proposed. Initially, gradient information and truncation coefficient are added to the Absolute Differences (AD) measure function to highlight the image edge, avoiding the excessive value caused by pixel outliers; Then, a fast Census transform method is proposed to work out the problem of high computational complexity of the traditional Census transform; Finally, the improved AD function constructs a cost calculation function incombination with the fast Census transform, which can strength for the robustness of traditional Semi Global Matching (SGM) algorithm in illumination noise and weak texture regions, and is capable to enhance the disparity map quality. Experimental results reveal that the average error matching rate of the improved approach is effectively reduced in comparison with traditional SGM matching algorithm, which settles the matter that the SGM algorithm at a single matching cost is arduous to obtain higher matching accuracy.