Fast CU Size Decision of Intra Prediction for Enhancement Layers in SHVC

Abstract

A fast CU depth level decision algorithm with low complexity for Scalable High Efficiency Video Coding (SHVC) is proposed to reduce the computational complexity caused by the recursive quad-tree structure of Coding Unit (CU) inherited from HEVC. The proposed scheme enables the early termination for CU sizes for intra prediction in the enhancement layer (EL) by using the property of the energy distribution in the current CTU and the optimal coding depths of the neighboring CTUs. Based on an adaptive thresholds scheme, the texture complexity can be categorized to three types and the CU depth level range is therefore narrowed correspondingly. Besides, the spatial correlation existing in the encoded neighboring CTUs is utilised to further reduce the number of Rate Distortion (RD) evaluation. The coding time is therefore significantly saved. The proposed algorithm is evaluated in the reference software of SHVC (SHM12.0). The experimental results shows that the reduction in encoding time can be up to 42.95% in comparison with the original encoding scheme in SHM.