Novel adaptive block-matching algorithm using modified switching search patterns

Motion estimation is regarded as vital importance for video coding. When designing search methods for motion estimation, both the estimation accuracy and computational complexity are the main considerations. The existing fast search methods such as TSS, NTSS, DS and ARPS can provide satisfactory result at a very low computational cost but within their own best application range. Besides, none of these methods can effectively jump out of the local optimum when processing large motion sequences. Therefore, in this paper, we propose a novel adaptive block-matching algorithm using modified switching search patterns. The proposed fast search method mainly adopts three techniques – adaptive threshold values for early termination of block-matching, block classification and local optimum sticking avoidance using ABCA – to remove unnecessary search, take full advantage of different search methods and avoid local optimum sticking problem. Experimental results showed that the proposed method has achieved significant improvement on both estimation accuracy and computational complexity.