The Research Of Transform Coding Techniques In Hybrid Video Coding Framework

Currently mobile phones and tablet PCs are turning2K (2580x1600) resolution. Television began to shift4K (3840x2160) and online video websites enhance the resolution of720P/1080P resolution. With the large popularity of HD video, network transmission speed and capacity of the storage medium is rapidly improved. But there is still demand on higher video compression efficiency. The release from last generation standard H.264has been ten years. In the past decade a lot of new algorithms proposed to provide support for the new generation of video compression standards. In April2010, the International Organization for Standardization ISO/EEC MPEG group and the ITU-T VCEG group co-founded JCT-VC group (Joint Collaborative Team on Video Coding), held its first JCT-VC in the German city of Dresden meeting, marking a new generation of digital video compression standard HEVC (High Efficiency Video Coding) began. In July2010, JCT-VC group released HEVC standard coding efficiency of the test platform candidate TMuC (Test Model under Consideration).16AD-HOC groups are established to receive and review the coding standard proposal. In October2010, JCT-VC group released the first HEVC standard coding efficiency test platform HM (HEVC test Model).In January2012, JCT-VC group released HEVC Committee for Standardization standard first draft CD (Committee Draft). In July2012, JCT-VC group released HEVC standard first draft international standard DIS (Draft International Standard). In January2013, JCT-VC group released HEVC standard first Final Draft International Standard FDIS (Final Draft International Standard). In April2013,, HEVC officially was accepted as the next generation video cosing standard with the name of H.265by ITU-T in Geneva.In this paper, the LTE system downlink packet scheduling technique has been deeply researched and two new scheduling algorithms are proposed based on the traditional proportional fair scheduling algorithm. Detailed theoretical analysis and simulation are completed. The main work is summarized as follows:First, we discuss the basic principle of video compression coding framework. We focuse on waveform coding. After analyzing H.261standard coding frame, encoding algorithms and stream three layered structure, three kinds of transform coding are given separately. H.261standard uses the original64-bit floating-point precision DCT. H.264standard uses integer DCT transform with adjustment at coefficients. HEVC uses the integer approximation DCT transform.Secondly, we proposed a prediction mode dependent directional DCT transform algorithm for HEVC standard. Then we discuss the HEVC standard development status, key technologies and compression performance. A large number of transformation algorithms emergent in HEVC standardization process, which are divided into two categories. One is with a combination of prediction models and transform, the other one is based on image content of the training transformation matrix. The proposed prediction mode dependent directional DCT algorithm (MD-DDCT) can be seen0.39%coding gain after the analysis of simulation results for the BasketballDrill sequence with obviously directional content.Thirdly, after study on the orthogonality and normalization of integer DCT for HEVC standard, we proposes integer DCT transform matrix with better orthogonality. After introducing the coefficients selection of integer DCT transform for HEVC standard, we gives the orthogonality judgment model and orthogonality verification method and proposes new integer DCT transform matrix. For lack of normalization, after analyzing the quantitative process, we give the normalization compensation matrix. As coding gain up to0.32%can be seen from the test results in a high resolution high rate case. In most cases, it does not achieve significant gain. The reason is that at lower bit rate, the quantization step is large and the orthogonality and normalization error introducted by integer DCT can not be revealed.Finally, after the analysis of sources of computational complexity of the HEVC standard, we give the history of Intel x86computing platform and SIMD instructions set for multimedia. Combined with the characteristics of SAD and DCT in HEVC standard, optimized assembly codes are given based on the SIMD instructions. Finally those assembly codes are embed to HEVC test model HM and take test according to the HEVC test specify. As can be seen based on SIMD instruction sets significant accelerations are obtained from the test results. In the calculation of4x4DCT,61times in acceleration can be obtained and in the calculation of8x8DCT,13times in acceleration can be obtained.