Skip to Main Content
A mufti-resolution technique, based on translation invariant discrete multi-wavelet transform, for matching a stereo pair of images is presented. The technique uses the well known coarse to fine strategy, involving the calculation of matching points at the coarsest level and refinement up to the finest level calculating dense 2D disparity maps. The technique addresses the estimation of optimal corresponding points and the corresponding disparity maps using local scale normalization at each decomposition level to deal with the illuminative variation between the pixels or the image pairs. The problem or ambiguity, and occlusion to some extent, is addressed by using a geometric topological refinement procedure which works with in the constraints of uniqueness and continuity.