Motion estimation using a compressive sensing architecture for H.264/AVC

Compressive sensing provided the probability of reconstructing signals using significantly fewer samples than the Shannon sampling theorem would normally permit. This encouraged people to try to collect less information other than to compress them during signal processing and reconstruction. The main focus in compressive sensing lies in the reconstruction of basis set and nonlinear optimization. This paper used a compressive sensing architecture to reduce the computation complexity in motion estimation of H.264/AVC, through constructing a measurement matrix to connect the reference macro block and current macro block. Then we achieve an analytical resolution and reduced computation cost without matching accuracy lose.