Reducing The Complexity Of Inter-prediction Mode Decision For High Efficiency Video Codec

Abstract

High Efficiency Video Coding (HEVC) standard is the latest joint video project of the ITU-T Video Coding Experts Group (VCEG) and the ISO/IEC Moving Picture Experts Group (MPEG) standardization organizations, working together in a partnership known as the Joint Collaborative Team on Video Coding (JCT-VC). While the HEVC is based on the same architecture of the widely used H.264/AVC standard [8], it includes many new coding tools and almost all the encoder blocks are optimized with respect to their counterparts in the H.264/AVC. This allows the new standard to achieve up to 50% bitrate reduction compared to its predecessor at the same visual quality at the cost of increase in complexity.Like H.264/AVC, mode decisions with motion estimation (ME) remain among the most time-consuming computations in HEVC. In an inter-prediction mode decision, a full-search algorithm searches for every possible block size and refines the results from integer-pel to quarter-pel resolution. Thus, a full-search algorithm guarantees the highest level of compression performance. However, the considerable computational complexity for a mode decision is critical for the encoding speed.In this thesis fast adaptive termination algorithm is proposed to terminate early the mode decision in inter-prediction for HEVC. Based on rate distortion (RD) cost, all the inter prediction modes are classified as skip or non-skip modes and to select the best mode minimum RD cost of these two modes are predicted. For skip mode, mode decision is predicted in early stage while in non-skip mode different stages are proposed to speed-up the mode decision. Experimental results based on several video test sequences suggest decrease of about 25%-40% in encoding time is achieved with implementation of Fast Adaptive Termination algorithm for inter-prediction mode decision with negligible degradation in peak signal to noise ratio (PSNR). Metrics such as BD-bitrate, BD-PSNR, SSIM and computational complexity are also used.