Application of Moment Invariant Approach in Region Matching

A method of region-based matching is presented in the paper. In addition to the traditional constraint conditions such as order constraint and unique constraint, a new constraint condition according to epipolar constraint is introduced to this algorithm where the relative position constraints better than adjacent constraint are adopted. This method can effectively handle binocular stereo matching in the case where occlusion regions and deformed regions are in stereo image pair. During selection of a best region among multiple candidate regions, the concepts on gravity distance and moment distance are proposed and used with moment invariant theory to determine the best matching region. The experimental results indicate that the approach using invariant moment to match stereo images is valid and feasible.