Using two-stage sorting-based partial distortion search for motion estimation in H.264/AVC

Abstract

The H.264/AVC video coding standard uses seven variable block sizes ranging from 16×16 to 4×4 in interframe coding. The motion estimation (ME) with seven modes needs very high computational complexity. To reduce the complexity of the ME module in H.264/AVC, the conventional partial distortion searching (PDS) algorithm, from a top-to-bottom sequential line-matching scan, is employed for fast full search (FFS) on the test software JM (called JM-FFS). We propose a two-stage sorting-based PDS (TS-PDS) algorithm for a fast lossless ME module in H.264/AVC by using the characteristics of pattern similarity matching errors. Two significant features of the image block pattern, bit-plane and absolute difference of means (ADM), are used to determine the calculation order in the TS-PDS. Moreover, a two-stage sorting-based normalized PDS (TS-NPDS) for a fast lossy ME module is also implemented in the same way. The TS-PDS can find the same motion vectors as the JM-FFS, and the TS-NPDS can obtain relatively accurate motion vectors with a large reduced computational load. Simulation results show that the average of the computational reduction ratio from the TS-PDS is 14.5% over the JM-FFS without any loss of peak signal-to-noise ratio of luminance (PSNRY) performance, and the TS-NPDS is 39.6% over the JM-FFS with the PSNRY performance very close to the JM-FFS.