Study On Theory And Algorithms For High-efficiency Video Coding

With the gradual merging of multimedia technology and network communication technology, there have been fascinating developments in video coding technologies during the past two decades, which gave birth to a series of video coding standards for different applications, notably the ITU-T H.26x and ISO/IEC MPEG-x families. The continuous renovation of video coding standards resulted in significant improvements of video coding efficiency. However, the basic coding schemes of these video coding standards are almost the same, namely block-based hybrid video coding. For the time being, the hot research topics of both industrial and academic circles are focusing on the four basic building blocks of hybrid coding scheme, namely predictive coding, transform coding, entropy coding and quantization, so an intensive study on the theory and technologies involved in these building blocks has great theoretical and practical significance. This paper makes an intensive study on the theory and algorithms for high-efficiency video coding, and the main contributions of this paper is listed as follows:Predictive coding theory is summarized, and sources of predictive coding gains are theoretically analyzed. After an introduction of the developing process of the intra and inter predictive coding technologies, this paper makes an intensive research of the latestly proposed advanced intra/inter predictive coding methods, and analyses their basic concepts and principles in details, highlighting problems and research directions.The theory for orthogonal transform in video coding is analyzed and summarized thoroughly in this paper, and a new performance measurement of transform coding is proposed. After an intensive study and detailed presentation of the Mode Dependent Directional Transform algorithm, a new scheme of texture-based adaptive transform is presented.Thoroughly research of the entropy coding theory and algorithms in video coding is made in this paper. First, it analyzes and summarizes the defects of Variable Length Coding algorithms in video coding. Second, it develops the arithmetic coding theory by proposing the theorem of the optimal binary probability model number in binary arithmetic coding applications, and gives a theoretical demonstration of the equivalent coding efficiency for binary arithmetic coding and multi-symbol arithmetic coding. The design principles of binarization and context modeling involved in binary arithmetic coding schemes are also analyzed and summarized in this paper. Third, a new entropy coding method of transform coefficients and a new texture-based adaptive entropy coding method of intra prediction modes are proposed in this paper.The rate-distortion optimization theory and algorithms, which are closely related to quantization, are researched in this paper. It makes a comprehensive introduction and analysis of Lagrange optimization algorithm, and analyzes the applications of Lagrange optimization in video coding. A block-joint rate distortion optimization scheme is proposed in this paper, and applied to the optimization of intra prediction mode decision.