Refinement of Inverse Depth Plane in Textureless and Occluded Regions in a Multiview Stereo Matching Scheme

Figure 1

An example of disparity map generation using an image with wide-textureless and occluded-textureless areas (Midd2 of Middlebury 2006 dataset). Conventional cost volume-based multiview stereo vision algorithms (PSS, Soft3D, and EnSoft3D) have limitations to refine the matching costs in the textureless areas. The proposed method can refine the matching cost in the textureless area and obtain very accurate disparity maps.