Low computational complexity variable block size (VBS) partitioning for motion estimation using the Walsh Hadamard transform (WHT)

摘要

Variable Block Size (VBS) based motion estimation has\udbeen adapted in state of the art video coding, such as\udH.264/AVC, VC-1. However, a low complexity H.264/AVC\udencoder cannot take advantage of VBS due to its power consumption\udrequirements. In this paper, we present a VBS partition\udalgorithm based on a binary motion edge map without\udeither initial motion estimation or Rate-Distortion (R-D)\udoptimization for selecting modes. The proposed algorithm\uduses the Walsh Hadamard Transform (WHT) to create a binary\udedge map, which provides a computational complexity\udcost effectiveness compared to other light segmentation\udmethods typically used to detect the required region.