Research On Rate-Distortion Optimization Techniques For HEVC Standard

With the rapid development of computing technology, communication technology and multimedia technology, multimedia applications have been pervasive in people’s daily lives. Video contains great amount of information, and is the most popularly used media among multimedias. However, due to the huge amount of video data, video cannot be transmitted in the Internet without compression. Video storage, transmission, and playback are premised on video coding technique, which is also the key enabling technique of digital television, video surveillance and network video applications. The video coding standard is the common basic standard for digital audio and video industry, and provides a unified technical specification and bit stream format for video compression. The video coding standard is very important to the development of the entire consumer electronics industry. Among the video coding standards, the latest video coding standard, High Efficiency Video Coding (HEVC), has already been issued in January2013.Rate-distortion optimization (RDO) technique is one of the key techniques in video coding, which is based on Shannon’s rate distortion coding theory and extensively used in video coding to strike a balance between bit consumption and video distortion. RDO aims to cost as few bits as possible but obtain as good video quality as possible so as to achieve the optimal rate-distortion performance. This dissertation focuses on the application of rate-distortion coding theory in video coding, especially on the application of RDO technique to the emerging international video coding standard, HEVC, to improve the coding efficiency and to control the bitrate more accurately.The main innovations and contributions of this dissertation are listed as follows.(1) This dissertation proposes a reference picture management strategy based on RDO technique. The dissertation models the reference picture management problem as a rate distortion optimization problem through deep analysis, and derives the optimal solution based on several reasonable assumptions. After determining the optimal referencing structure, the dissertation addresses the picture quality adjustment problem to further improve the coding efficiency, which is still based on the RDO technique. Besides, the dissertation develops several fast algorithms for the reference picture management problem, which is applicable for the applications which require low complexity and low delay encoding. Part of the reference picture management methods proposed in the dissertation have already been adopted by the latest international video coding standard HEVC, and has already been integrated into the HEVC reference software.(2) This dissertation proposes a Quantization Parameter (QP) refinement algorithm based on RDO technique. To overcome the complexity introduced by multiple-QP optimization, the dissertation analyzes the relationship between the optimal QP value and the Lagrange multiplier used in the RDO process, based on which QP refinement algorithm is proposed. When the QP refinement algorithm is applied, without increasing the encoding complexity, similar coding efficiency improvement can be achieved as5-QP optimization (5-QP optimization introduces additional4times encoding complexity). The proposed QP refinement algorithm has already been adopted by the latest international video coding standard HEVC, and has already been integrated into the HEVC reference software.(3) This dissertation proposes a λ-domain rate control algorithm based on RDO technique. After analyzing the difficulties of applying the traditional R-Q (Rate-Quantization) model based rate control algorithm to HEVC, this dissertation builds an R-λ model, which better reflects the rate-distortion relationship. Furthermore, the dissertation proposes an R-λ model based rate control algorithm, and applies it to the latest video coding standard HEVC. Compared with R-Q model based rate control algorithm, R-λ model based rate control algorithm not only achieves both better rate control accuracy and better rate-distortion performance, but also achieves more accurate instant bitrate. The R-λ model based rate control algorithm proposed in this dissertation has already been adopted by the latest international video coding standard HEVC, and has already been integrated into the HEVC reference software.