Fast Encoding Algorithms for SHVC Intra/Inter Coding.
Xin Lu,Chang Yu,Graham Martin
DOI: https://doi.org/10.1109/dcc.2019.00107
2019-01-01
Abstract:Scalable High-efficiency Video Coding (Scalable HEVC, SHVC) is the most recent scalable video coding standard, which is developed by the Joint Collaborative Team on Video Coding (JCT-VC). In order to meet the needs of various client terminals, network conditions, and user demands, a multi-layer encoding framework is employed to generate the scalable bitstream. In addition to the desirable scalabilities, SHVC also aims to achieve better coding efficiency than previous standards. Therefore, all the advanced coding tools ofHEVC are inherited, and several additional tools are specifically designed to support scalability. However, the superior coding efficiency is achieved at the cost of significantly increased computational complexity. Given the fact that an SHVC encoder is considerably more complex than an HEVC encoder, this makes a great demand of a fast SHVC encoder that incurs only a negligible degradation in compression efficiency. In order to address the above issue, several fast schemes have been suggested for SHVC intra and inter coding. However, temporal, spatial and inter-layer correlations have not been fully exploited. In addition, most existing algorithms aimed only at the improvement either on the intra coding process or the inter coding process. If all the available information including the texture complexity, the motion activity, as well as different types of correlation can be appropriately utilised, it is expected that the computational complexity can be reduced further. Therefore, a fast Coding Unit (CU) size decision algorithm is proposed for spatial SHVC intra coding to early terminate the CU size decision process. The energy distribution is exploited to classify the texture complexity, and the available information from the BL is used to narrow the coding depth level ranges for the LCUs in the EL. Unlikely intra modes are removed by considering the inter-layer, spatial and temporal correlations. Different candidate mode lists are constructed for the PUs of various sizes. In addition, the implementation after RMD is also improved. A fast Prediction Unit (PU) prediction mode decision method is also suggested for SHVC inter coding, in which the motion complexity and the inter-layer dependency of PU modes are jointly used to determine the best PU prediction mode for the PU in the EL. The proposed algorithms were implemented in the SHVC reference software SHM 12.0. Two-layers encoding structure was employed for spatial scalability, and the scalability factor was 2. Common test conditions were satisfied. The All Intra (AI) and Lowdelay configurations were employed to evaluate the proposed fast intra and inter coding algorithms. The quantisation parameters (QPs) for the BL and EL were (22,20), (27,25), (32,30) and (37,35). Experimental results showed that the proposed fast intra algorithm reduces the encoding time by 48% on average, while the proposed fast inter algorithm achieves an average time reduction of more than 36% over the unchanged SHM encoder, and the loss in Rate-Distortion (RD) performance is acceptable. It can be concluded that the proposed algorithms outperform the algorithms previously proposed.