Examination Committee

Prof Kai Lung HUI, ISOM/HKUST (Chairperson)
Prof Ling SHI, ECE/HKUST (Thesis Supervisor)
Prof Hongkai XIONG, School of Electronic Information and Electrical Engineering, Shanghai Jiao Tong University (External Examiner)
Prof Bing ZENG, School of Electronic Engineering University of Electronic Science and Technology of China (External Examiner)
Prof Danny Hin Kwok TSANG, ECE/HKUST
Prof Wai Ho MOW, ECE/HKUST
Prof Gary Shueng Han CHAN, CSE/HKUST

Abstract

The Joint Collaborative Team on Video Coding (JCT-VC) has recently accomplished a new standard, referred to as high efficiency video coding (HEVC) whose primary goal is to achieve 50% bit-rate reduction under an equal perceptual quality as compared to its predecessor. However, there is a constant hunger for higher quality video compression algorithms because the amounts of images and videos are now growing explosively. Therefore, what I focus on in this thesis is further improving video coding efficiency based on HEVC.
 
First, I present a new block-based method for the HEVC intra coding. Pixels in each prediction block are divided into two parts: half pixels are coded via a novel padding technique together with a constrained quantization algorithm whereas the other half are reconstructed by linear interpolations. Second, I propose a new adaptive sharpening filter based on guided image filter and embed it between deblocking filter and SAO. The proposed algorithm classifies pixels of a frame into several groups according to each pixel’s Sum-Modified-Laplacian value and assigns identical optimal filtering parameters to the pixels belonging to the same group based on rate-distortion optimization. Third, instead of the default transforms in HEVC, I build an elliptical model with directionality and design some non-separable transforms based on KLT in closed-form for each intra-prediction mode. Fourth, I present a DC coefficient estimation algorithm for intra-predicted residual blocks, which solves a pixel domain optimal offset in a closed-form to recover the corresponding block edges based on the texture continuity priori hypotheses.
 
Simulation results demonstrate that the overall performance can achieve 3.5%, 2.4%, 2.3% and 2.3% BD-rate reductions on average under AI, RA, LDP and LDP configurations, respectively, with slight complexity increase. For some specific sequences, the coding efficiency enhancement can be up to 10.2% and 8.8% under AI and LDP configurations, respectively.