3D correspondences in textured depth-maps through planar similarity transform

Abstract

Many low cost realtime depth-map reconstruction devices recently appeared on the market opened new opportunities for the Computer Vision community to integrate these information in many research areas. The knowledge of the underlying depth-map together with a visual snapshot of the scene can greatly improve the robustness of points matching between different views even for wide baseline acquisitions. In this paper we are presenting how visual correspondences from different views can be identified by robust Similarity Invariant Descriptors (SID) once their laying plane is known. Furthermore depth-map, providing with a rough geometrical description of the underlying scene, allows to select only feature points belonging to almost planar regions, skipping geometrical corners or edges that undergo non-linear distortion for viewpoint changes. The proposed SIDs keep much more information of the original area with respect to commonly used affine invariant descriptors, like SIFT of GLOH, making the proposed approach much less prone to false matches even for wide viewpoint changes.