Dense stereo matching in depth discontinuity areas based on edge detection and nearest valid point search

Dense stereo matching based on two viewpoint images is an important step in 3D reconstruction. In this study, a new disparity estimation algorithm for depth discontinuity areas proposed. implemented the basis of semi-global algorithm. If ground truth known, then mismatched pixels are selected as those with difference more than one pixel from truth. Otherwise, left-right consistency checking process used. Then, Canny operator used to detect image edges. For each pixel, maximum total eight nearest edge points exists four pairs symmetric directions. Basically, searching valid point proceeds along directions following sequence: horizontal, vertical, 45° oblique and 135° oblique. Specifically, pair directions, searched direction distant point. successfully reassigned by found, terminated. Comparative results obtained using well-known benchmark datasets unmanned aerial vehicle demonstrate excellent performance proposed method.