Video Coding with R-D Constrained Hierarchical Variable Block Size (VBS) Motion Estimation

Abstract

The variable block size (VBS) motion estimation technique allows for larger blocks to be used when smaller blocks yield little gain, saving bit rates especially for areas containing more complex motion. However, the employment of the VBS motion estimation technique introduces a new optimization issue for the motion compensated transform coding, because an increase in bit rate allocation is necessary to accommodate VBS motion vectors. Hence, this paper will describe a rate-distortion (R-D) constrained approach for the hierarchical quadtree VBS motion estimation and displaced frame difference (DFD) coding. First, in order to make the R-D tree search feasible, the hierarchical quadtree VBS motion structures are grouped into two-stage model structures coupled to an efficient R-D tree search method. Next, a R-D constrained solution based on Lagrange multiplier method controls the employed VBS motion vectors in relation to the dependent DFD coding. Extensive computer simulation employing the motion compensated transform coding technique shows that an overall improvement up to 1.0 dB is obtainable, compared to the fixed block size motion estimation.