Adaptive multiple-candidate hierarchical search for block matching algorithm

A new adaptive multiple-candidate hierarchical search for the estimation of motion vectors is presented. This technique can eliminate the undesirable problem of local minima in coarse-to-fine hierarchial block matching algorithms. Computer simulations show that the algorithm is very efficient as compared with the traditional coarse-to-fine hierarchial search. Also, it is close to the performance of the exhaustive full search, with a significant reduction on computation.Department of Electronic and Information Engineerin