Efficient search algorithms for block-matching motion estimation

In this paper, two new and efficient algorithms for block motion estimation are presented. The proposed algorithms concentrate on improvements to existing block motion estimation algorithms (BMAs) and include four effective steps: prediction of static blocks, prediction of optimal search starting point, motion categorization for the block and motion vector search around the optimal starting point. Two different search patterns are presented that achieve different tradeoffs between search speed and reconstructed picture quality. Extensive simulation results and comparative analysis with the well-known Diamond search show that the proposed algorithms outperform conventional algorithms in terms of computational complexity while maintaining picture quality.