Research On Fast Mode Decision For Intra Frame Based On HEVC

With the popularization and development of computer technology, multimediatechnology makes continuous improvement. Multimedia technology consists ofcomputer technology and video technology which occupies half of the country. As thelarge resolution of intelligent TV brought into thousands of households, users havehigher needs for high resolution videos. The current video coding standard can notsolve this problem. In order to acclimate the upcoming HD time request, twoorganizations--ITU-T VCEG and ISO/IEC MPEG--issue a new generation codingstandard jointly called High Efficiency Video Coding (HEVC). HEVC uses newcoding framework and coding technology to obtain higher coding performance.Coding process can consume lots of time, especially in the prediction stage. In thiscase, the encoder can cost several days to coding a video sequence, which is difficultto be used in our life. Therefore, how to reduce the encoding complexity withoutperformance loss has become an imminent task.This thesis studies on the Rough Mode Decision (RMD) process and fast codingunit size decision based on the HEVC intra prediction process and improves twoprocesses to reduce the encoding complexity.First, the RMD process has several aspects which need to be improved. Thisthesis plans to use the candidate list information of parent prediction unit and currentprediction unit to reduce the number of modes in candidate list. In addition, neighbormodes which have a large difference in candidate list cost have low possibility to bethe best mode. Based on this fact, we propose a method to further reduce predictionmodes. Then, we make full use of the relationship of35modes in intra prediction bygenerating several modes groups. In the end, we proposed a combination algorithm toreduce the number of modes, which may be checked in RDO process.Second, in the process of determining the coding unit size, this paper madestatistics on adjacent coding unit size and we analyze the data to get the conditionalprobability which can determine the current coding unit size. In addition, we can usethe conditional probability to distinguish between non optimal size coding unit andthe optimal size coding unit. We can check more prediction modes in optimal sizecoding unit and less prediction modes in non optimal size coding unit. So we cansimplify the coding process to reduce the coding complexity.In conclusion, we improved the HEVC intra prediction process in two aspects.The results show that the proposed algorithm has1.1%loss in the encoding quality, and save the encoding time by35%.