Fast motion estimation in Walsh Hadamard domain

In this paper, we propose a fast block-based motion estimation algorithm for video processing using Walsh Hadamard search (WHS) approach. The proposed algorithm performs motion estimation by projecting blocks on Walsh Hadamard (WH) basis picture, which allows elimination of mismatch candidates on an early stage of the matching process, therefore avoid excessive operations. For the remaining candidates, we propose a block pyramid matching to evaluate the similarity between blocks to further reduce the computation requirement. Experimental results show that the computation of the proposed algorithm is about 5% of traditional full search block matching (FSBM), while the mean square error is close to FSBM.