VLC Based On Second-run-level Coding And Dynamic Table Truncation

With the increasing demand of multimedia service, video coding technology becomes more and more attractable and important in multimedia application. Nowadays, Block-based hybrid video coding scheme has been widely applied as an efficient coding solution and has been adopted as the basic coding framework in H.26x and MPEG standards. Entropy coding is a crucial part in the coding framework, whose compression efficiency is related directly to the output bitrate of the entire video coding system. The methods of entropy coding can be classified into two categories as variable length coding (VLC) and arithmetic coding (AC). In the latest National Audio and Video Coding Standard (AVS), the residual coefficients are coded by the method called Context-based 2D-VLC. Compared with other advanced VLC techniques adopted by international video standards, the coding efficiency of this method is not so satisfactory. Supported by the National High Technology Research and Development Program of China (863 Program)–Joint Source-Channel Coding (No: 2002AA119010), this thesis focuses on the novel VLC method with good performance and independent intellectual patents.After introduction of basic video-coding theory and analysis of the entropy coding techniques in mainstream video standards, a novel VLC method designed for residual coefficients is proposed in this paper. This method exploits and utilizes the context information sufficiently and achieves much better compression efficiency, by employing three novel approaches-2nd-run-level coding, dynamic table truncation and adaptive codes selection. After traditional run-level coding, a large amount of statistical redundancy still exists in the level sequence and run sequence, thus second run-level coding is first introduced to compress these levels and runs more efficiently. Different from the prediction information utilized by most adaptive coding techniques, another valuable context information: source constrained information is defined in this thesis. In order to achieve a higher degree of context adaptability and coding efficiency, a method called dynamic table truncation that utilizes the source constrained information sufficiently is proposed and employed in the sequential coding of 2D symbols, without large amounts of VLC tables required. In addition, the coding performance can be further improved by enabling adaptive codes selection without increasing any computational complexity. Experimental results show that when compared with context-based 2D-VLC, the novel VLC gains 0.25 dB ~ 0.79 dB in PSNR and achieves 5.30% ~ 11.58% improvement in bit rate reduction.In this paper, a novel VLC method based on second run-level coding and dynamic table truncation is proposed to compress the residual coefficients more efficiently. Relative to Context-based 2D-VLC adopted in present AVS standard, the novel VLC performs much better in compression efficiency. Therefore, this method is recommended for video coding system to reduce its output bitrate and improve its coding performance.