Image Edge Based Efficient Stereo Matching
Zhiguang Zhang, Jianhua Zou · 2019
In this paper, we present the image-edge based stereo matching method-an extension of the ELAS algorithm- which increases the robustness of stereo matching. The image region is divided into triangular regions by triangulation on support points. Based on image Canny edge information, our approach divides triangular regions into three groups according to whether piecewise disparity smooth is valid or not. The Bayesian method is used for piecewise disparity smooth areas, and the local stereo method is used for disparity discontinuity areas. For the disparity smooth areas, we further strengthen the influence of disparity prior by making use of color prior. Based on experimental results, our method is comparable in accuracy and efficiency.