An efficient inter-frame coding with intra skip decision in H.264/AVC

In this paper, we propose a simple but effective intra-mode skip algorithm to reduce the computational cost for inter-frame coding. It makes use of motion, temporal, and spatial homogeneity characteristics of video sequences. Specifically, the motion homogeneity is defined by using the mean deviation of motion vectors (MV) of the 4x4 blocks. The temporal and spatial homogeneity are computed by the sum of absolute difference (SAD) between an original block and its prediction block of the best inter-mode and the intra-mode with large block size, respectively. Based on the three types of homogeneity, the proposed method skip the full intra-mode search for inter-frame coding to reduce the encoding time when a region has the motion homogeneity and temporal homogeneity is stronger than the spatial one. Experimental results demonstrate that the proposed algorithm significantly increases the skip decision accuracy by up to 20% and reduces the total encoding time by about 3~10% as compared to the existing methods with negligible loss in PSNR and small increment of bit rate.