Research On Fast Intra Prediction Algorithm For High Efficiency Video Coding

With the increasing improvement of multimedia technology, high-definition television is more and more popular so that video data is also increasing simultaneously. However, being difficult to transmit the giant video in the practical application, the demand of high efficiency video coding is growing more and more urgently. Therefore, on the basis of H.264 being unable to meet the increasing demand, High Efficiency Video Coding is drawn up by the Video Coding Group (JCT-VC) formally. However, with the nearly double augment of the compression efficiency, the fact which the computation complexity is added by almost two to four times brings hugely fatal damage to the coding speed. Therefore, how to reduce the algorithm complexity effectively, speed up the speed of video compression, and promote the development become the key to popularization and commercialization of HEVC application in the condition of ensuring video quality.By guarantee the acceptable quality and reducing algorithm complexity, the thesis researches relatively time-consuming modules and analyzes the cause of algorithm complexity in the HEVC. Then we raise two fast algorithms, which are the rapid prediction algorithm based on SSIM and MPM and the fast mode judgment algorithm based on probability of candidate patterns. Their main directions include of both adding a feature to skip the existing algorithm and using numerous strategies to reduce the number of candidate prediction models, so that computational complexity is optimized on the premise of guarantee performance.On the basis of both each prediction block finishing rough mode decision by cost function of SATD and characteristics between SSIM & MPM and optimal prediction model, the thesis raises the rapid prediction algorithm based on SSIM and MPM so that not every prediction block is achieved rough mode decision and unnecessary rate distortion optimization is avoided.According to the problem of different PU size having too many candidate models and modes of different cost value having inequality probability of selecting as the optimal prediction model, the thesis raises the fast mode judgment algorithm based on probability of candidate patterns so that the number of candidate models are reduced. Then video test sequences of the Class A to Class E are used to contrast both the two algorithms with HEVC, the experimental results show that the algorithms could reduce encoding time of HM12.1 by 31.46 percent to 46.37 percent on the basis of acceptable video quality and compressed rate having no obvious change. Moreover, by the both the curve of RD and subjective quality of reconstruction image, there is almost no difference between the proposed algorithms and HEVC. Therefore, the two algorithms have practical value in the practicability and commercialization.