A stereo matching algorithm based on probability relaxation is presented. The algorithm combines feature-based and area-based relaxation for matching images of real scenes and takes the spatial relationships between a central point and its neighbouring points into account. Using iterative refinement of the probabilities, the matching process can reduce ambiguity and noise, and select the best match among several possible choices. Test shows that the algorithm has sub-pixel matching accuracy for the normal image pair.
展开▼