Research Of Fast Coding Techniques For SHVC

With the sustainable development towards high definition and high frame rate applications, the new generation video coding standard,HEVC, and its scalable extension,SHVC are making full efforts in their standardization. As the next generation video coding standard, HEVC have significant coding efficiency advantage over all existing ones. However, the complexity of HEVC codec is extreamly heavy, and its scalable extension, SHVC, has even higher complexity. Therefore, reducing the complexity of the SHVC has become one of the hot research direction with negligible quality reduction. In this thesis, three different fast encoding algorithms aiming to HEVC quality scalable video coding and spatial scalable video coding are proposed to deal with the problem. The main work and innovation are as follows:(1) A fast encoding algorithm was proposed to reduce the complexity of enhancement layer, which based on the corresponding depth information of base layer. This algorithm based on the correlation between base layer and enhancement layers. The main idea of the algorithm is that during the enhancement layer coding, each enhancement layer coding tree unit refer to its corresponding actual coding depth of base layer as its largest unit coding respectively. Thereby, this algorithm can reduce unnecessary calculation process, and reduce the complexity of the coding. Besides, in order to further reduce the complexity of encoding, part of the process of mode selection was skiped.(2) For quality scalable video coding, a fast encoding algorithm aims to reduce entire coding complexity of encoder was proposed, which based on the reference frame and base layer encoding depth information. This algorithm based on the correlation of reference frame and current frame as well as the correlation of base layer and the enhancement layer. The main idea of the algorithm as follows. When base layer of the current frame is encoded, current frame coding tree unit refer to its corresponding actual coding depth of reference frame as its largest unit coding respectively, thereby reducing the current frame unnecessary mode selection and the process of rate-distortion cost calculation, and thus reducing the complexity of the coding. For the enhancement layer coding, each enhancement layer coding tree unit refer to its corresponding actual coding depth of base layer as its largest unit coding respectively, thereby this algorithm can reduce unnecessary calculation process, and reduce the complexity of the coding.(3) For quality and spatial scalable video coding, we propose a fast encoding algorithm based on the above studies to achieve quality scalability, and this algorithm supports spatial scalability in any proportion. The main idea of the algorithm as follows. For quality scalable, when coding base layer reference its previous frame coded depth information and when coding enhancement layer refer to its corresponding base layer coded depth information. For spatial scalable coding, when coding base layer refer to its previous frame coded depth information and when coding enhancement layer refer to its corresponding base layer coded extended depth information. Thereby, this algorithm can reduce the complexity of the entire encoder.To verify the performence of proposed algorithms, the thesis provides detailed experimental results and analyses. Finally, the conclusion of this thesis is given, as well as the prospect of the further work.