A Two-Stage Fast CU Size Decision Method for HEVC Intracoding.

Qi Zhang,Jun Sun,Yizhou Duan,Zongming Guo
DOI: https://doi.org/10.1109/mmsp.2015.7340816
2015-01-01
Abstract:In HEVC coding standard, the encoder employs a quad-tree-based Coding Unit (CU) structure to adapt to various texture characteristics of images. Although it provides better compression performance, the computation load increases drastically. In this paper, a two-stage fast CU size decision method is proposed for HEVC intra coding. In the first stage, we utilize the combination of the weighted variance and the maximum value of the Hadamard costs of the four subCUs to represent the CU complexity, and classify each CU into three categories: compound, homogeneous, and undetermined. The former two kinds of CUs will be made early split and pruned decisions respectively. In the second stage, an improved SATD-based estimated R-D cost is employed to decide whether the prev undetermined CUs should be early pruned. Experimental results demonstrate that compared with the reference software HM13.0, the proposed fast CU size decision method provides 49% time reduction with slight quality degradation using the HEVC all intra test condition. This proposal has been adopted to provide HEVC video and picture encoding services at the server-side for UC mobile browser which covers over 100 million people in China.
What problem does this paper attempt to address?