Fast Coding Tree Unit depth decision for high efficiency video coding
Zhaoqing Pan, Sam Kwong, Yun Zhang, Jianjun Lei, Hui Yuan
Abstract
Zhaoqing Pan, Sam Kwong, Yun Zhang, Jianjun Lei, Hui Yuan
Abstract
High Efficiency Video Coding (HEVC) is the latest video coding standard, which adapts quadtree structure based Coding Tree Unit (CTU) to improve the coding efficiency. In HEVC encoding process, the CTU is recursively partitioned into coding units according to the quadtree depth. This technique increases the coding efficiency of HEVC, however, the achieved coding efficiency comes at the cost of high computational complexity. In this paper, we propose a fast C-TU quadtree depth decision algorithm to reduce the computational complexity of HEVC. Firstly, based on the best C-TU depth correlation among spatial and temporal neighboring CTUs, an early quadtree depth 0 decision algorithm is proposed. Then, according to the correlation between the prediction unit mode and the best CTU depth selection, a quadtree depth 3 skipped decision algorithm is proposed. Experimental results show that the proposed algorithm can achieve 40% on average encoding time saving, while maintaining a comparable rate-distortion performance.
OpenAlex reports 14 citations for this work. Citation counts describe recorded attention and do not establish research quality.
A contribution statement is not available in the OpenAlex record.
Method details are not available in the OpenAlex metadata.
Findings are not separately available in the OpenAlex metadata.
Limitations are not available in the OpenAlex metadata.
Application details are not available in the OpenAlex metadata.
High Efficiency Video Coding (HEVC) is the latest video coding standard, which adapts quadtree structure based Coding Tree Unit (CTU) to improve the coding efficiency. In HEVC encoding process, the CTU is recursively partitioned into coding units according to the quadtree depth. This technique increases the coding efficiency of HEVC, however, the achieved coding efficiency comes at the cost of high computational complexity. In this paper, we propose a fast C-TU quadtree depth decision algorithm to reduce the computational complexity of HEVC. Firstly, based on the best C-TU depth correlation among spatial and temporal neighboring CTUs, an early quadtree depth 0 decision algorithm is proposed. Then, according to the correlation between the prediction unit mode and the best CTU depth selection, a quadtree depth 3 skipped decision algorithm is proposed. Experimental results show that the proposed algorithm can achieve 40% on average encoding time saving, while maintaining a comparable rate-distortion performance.
Key concepts: Quadtree, Coding tree unit, Coding (social sciences), Computer science, Algorithmic efficiency, Computational complexity theory, Algorithm, Context-adaptive binary arithmetic coding