Research On Intra And Inter Prediction Techniques In Video Coding

In the era of big data,multimedia technology has achieved rapid development,the demand for diversification and high definition of video applications has gradually increased.In response to the challenge of video compression technology,Joint Video Exploration Team(JVET)was set up to explore the development of next-generation video coding standard-H.266.H.266 still follows the hybrid framework of mainstream video coding standards.At the same time,it introduces its own features such as more flexible quadtree and binary partitioning(QTBT),finer intraframe angle prediction direction,inter prediction modes suitable for special scenes(affine motion compensation prediction,etc.),adaptive multicore transform technology.This thesis studies intra prediction in three aspects and inter prediction in two aspects on H.266 respectively.The concrete work is as follows:Optimization of intra prediction techniquesFirst of all,if the coding block contains complex texture or the correlation between the reference pixels and the coding block is not very strong,the existing intra prediction method will produce serious distortion.To slove the problem,a new intra-prediction model based on template matching is proposed.The experimental results show that the proposed scheme can alleviate the distortion and improves the compression efficiency.Secondly,for the Planar mode,it is possible that the pixels in the boundary of the coding block and the neighboring coded blocks are discontinuous.A boundary filter is researched.This method can further improve the boundary pixel discontinuity problem.Finally,due to the fact that MPM mode is more likely to be choosed as the best mode,a method based on MPM filtering preselection mode is researched.This method reduces the coding complexity when the coding performance loss is negligible.Optimization of inter prediction techniquesFirstly,in modern video codecs standard,the bi-prediction signal is formed simply by averaging two uni-prediction signals using a weight value equal to 0.5.However,it will produce serious distortion,when illumination changes rapidly from one reference picture to another or the prediction quality of one motion-compensated prediction block may differ from the other due to factors such as quantization.Therefore,an adaptive bi-prediction method based on reference quality is proposed.Under LDB and RA configuration,the performance of this method is improved to a certain extent.Secondly,for the merge mode,a distance-based combination prediction method is proposed to improve the prediction accuracy of the coding block.The experimental results verify that the encoding performance is improved in the LDB,LDP and RA configurations.Among them,about 1.4% BD-rate saving can be achived on average with the LDP configuration.