Frame memory bandwidth reduction through dynamic search range update based on embedded compression ratio

Abstract

The H.264/AVC outperforms previous video coding with many tools which can increase R-D performance, but the memory bandwidth also increases remarkably. As the demand of high quality image goes up, a high frame resolution and bi-directional frames (B-frame) coding become essential. As a result, the memory bandwidth and memory access cycles are the biggest problem for a real time encoder. In a general tendency, the embedded compression (EC) is employed in the high performance encoder to solve this problem. In this paper, we explain the relationship between EC ratio and motion vector (MV), which can be used to reduce the memory bandwidth. Then, we proposed novel methods that uses dynamic search range (DSR) adjustment aware of EC ratio, in order to reduce a memory bandwidth and get a constant memory bandwidth with a lossless EC. We can guarantee an encoder to reduce till the 30% of the total memory bandwidth requirement with little image degradation.