Multi-baseline based texture adaptive belief propagation stereo matching technique for dense depth-map acquisition
Jin-Hyung Kim, Jae Wan Kwon, Yun-Ho Ko · 2014
In this paper a new multi-baseline stereo matching framework based on a modified belief propagation algorithm is presented to acquire dense depth-map. We propose a new matching cost, Extended Mean of Absolute Differences as local evidence in order to consider all possible disparity candidates and obtain dense depth-map. Also we propose a method that decides the weight parameter λ in belief propagation algorithm adaptively to local texture activity.