Two Stages Stereo Dense Matching Algorithm for 3D Skin Micro-surface Reconstruction

摘要

As usual, laser scanning and structured light projection represent the optical measurement technologies mostly employed for 3D digitizing of the human body surface. The disadvantage is higher costs of producing hardware components with more precision. This paper presents a solution to the problem of in vivo human skin micro-surface reconstruction based on stereo matching. Skin images are taken by camera with 90mm lens. Micro skin images show texture-full wrinkle and vein for feature detection, while they are lack of color and texture contrast for dense matching. To obtain accurate disparity map of skin image, the two stages stereo matching algorithm is proposed, which combines feature-based and region-based matching algorithm together. First stage a triangular mesh structure is defined as prior knowledge through feature-based sparse matching. Region-based dense matching is done in corresponding triangle pairs in second stage. We demonstrate our algorithm with active skin image data and evaluate the performance with pixel error of test images.