A Spatiotemporal Content-Based CU Size Prediction Method for HEVC

Abstract

The high efficiency video coding (HEVC) standard is a novel algorithm of video compression Different from the superior compression algorithm which is advanced video coding (AVC) the flexible partitioning scheme and increased prediction modes of HEVC provides better efficiency for encoding For this reason HEVC improves almost 50% in compression ratio However HEVC brings more computational complexity with the enhancing of compression performance and spends much more time on encoding Therefore we need an efficient method to reduce the encoding time while maintaining high compression ratio and less Bj?ntegaard delta bit rate (BDBR) loss In conventional HEVC encoder coding units (CUs) in a coding tree unit (CTU) are traversing from top to bottom in recursion This process is applied for determine the optimal splitting mode and occupies a lot of time in encoding In this dissertation we provide a fast CU decision to skip the unnecessary calculation The architecture of the proposed method includes adaptive depth-range prediction (ADRP) SAD-based region inspection (SRI) and texture-aware splitting termination (TST) Experimental results show that the proposed method can speed up 60 89% in low-delay and 66 77% in random-access compared to HEVC original encoder with 0 41 and 0 98 BDBR penalty meanwhile accompanying with -0 01 and -0 02 BDPSNR penalty

Date of Award	2017 Jul 22
Original language	English
Supervisor	Pei-Yin Chen (Supervisor)