Study On Optimization Of Intra Prediction Technologies In HEVC

HEVC（High Efficiency Video Coding standard）is the newest video codingstandard developed by the Joint Collaborative Team on Video Coding (JCT-VC),which is formed by ITU-T VCEG and ISO/IEC MPEG. Compared with the previousvideo coding standards, HEVC employs a lot of new techniques and thus has muchhigher coding efficiency. However, the improvement of its coding performance is atthe expense of heavy computational complexity. Therefore, how to reduce itscomputational complexity while still maintaining almost the same coding efficiencyhas become a crucial problem in a practical HEVC codec realization. Consequently,study on HEVC optimization has become a hot research topic and is also the mainconcern of this dissertation.To effectively reduce the computational complexity of HEVC, HEVC is firstlyintroduced in detail and its key techniques are thoroughly analyzed.Then,theoptimization algorithms for the Intra Prediction in HEVC are proposed in thisdissertation.First, by analyzing the charicateristic of intra prediction in HEVC, a novel fastCU decision algorithm for intra prediction is proposed based on the codinginformation of adjacent CU and a pre-determined threshold for speeding up theencoding process. This algorithm utilizes the depth used in adjacent CUs to estimatethe range of the current CU’s depth. By comparing the RDcost with a pre-determinedthreshold, an early termination scheme is further exploited for those CUs with N×Nsize or those boundary CUs.Second, a novel fast mode decision algorithm for intra prediction is presentedbased on direction and spatial correlation. This algorithm selects those more possibleprediction modes based on the evaluation of SATD as the candidate modes, which arethen classified into two classes: directional modes and non-directional modes. Thosedirectional candidate modes are further reduced by considering the optimal mode ofup layer’s PU and the MPM mode.As a result, the computational load is decreased dramatically.Finally, a novel fast intra prediciton algorithm is suggested by jointly using theproposed fast intra CU decision method and the fast intra mode decision method.Experimental results have shown that1) the proposed fast intra CU algorithmreduces43%computational complexity on average, while incurring only0.10dB lossin BDPSNR and1.78%increment in BDBR,compared with that of exhaustive CUsize decision in HEVC;2) the proposed fast intra mode decision algorithm canachieve a reduction of25%computational complexity on average, while incurringonly0.04dB loss in BDPSNR and0.50%increment in BDBR,compared with that ofexhaustive intra mode decision in HEVC;3) the proposed fast inta predictionalgorithm is able to obtain a reduction of55%computational load on average, withonly0.13dB loss in BDPSNR and2.35%increment in BDBR,compared with theoriginal HEVC intra prediction. In summary, all the fast algorithms proposed in thisdissertation are very meaningful to the application of HEVC.