Dense Stereo Matching by Hierarchical Belief Propagation based on Fuzzy Confidence Approach
FAKHFAKH ; KHOUDOUR ; EL-KOURSI ; BRUYELLE ; DUFAUX ; JACOT
Type de document
COMMUNICATION AVEC ACTES INTERNATIONAL (ACTI)
Langue
anglais
Auteur
FAKHFAKH ; KHOUDOUR ; EL-KOURSI ; BRUYELLE ; DUFAUX ; JACOT
Résumé / Abstract
One of the main challenges of stereo vision is to provide an accurate disparity map dealing with the occlusion and depth discontinuity problems. We propose a stereo matching algorithm using a hierarchical belief propagation method based on a selective approach. A confidence measure is computed for each pair of matched pixels. This is a new way of classifying the matched pixels into classes according to their confidence measure. The performances of the algorithm is evaluated on a standard dataset of the Middlebury stereo benchmark. The proposed algorithm proves its effectiveness compared with other the state of the art algorithms.
