Edge Traction Stereo Matching Method Using RBF

For obtaining the precise and dense disparity map, a stereo matching method with occlusion detection using edge traction was presented, which is integrated with the area similarity and edge feature points in radial basis function (RBF). Obtain the edge feature points by using LoG and their label sets of disparity to dividing the image region into feature and featureless sets. Area similarity, feature points and disparity smoothness are incorporated into error function of RBF. Relative reliable disparities of feature points restrict the disparity estimation of featureless points and decrease the possibility of forming mismatching illusion. Hierarchical Gaussians approach is imposed in this matching framework. Besides, we test the match values and the match conflict at sight lines to determine the occlusion area and avoid mismatching illusion. According the uniqueness constraint, whole high-density disparity map was obtained