Research On Low-complexity Encoding Techniques For The Scalable Video Coding Standard SHVC

Video has become the primary medium for obtaining and transmitting information in prople's daily life.However,the massive data of video leads to dramatical increase in encoding time.Due to the heterogeneity of the network and the differences between electronic terminals,the previous scalable video coding standards cannot achieve a real-time transmission,especially through the wireless network.To this end,the highly efficient scalable video coding standard H.265/SHVC,namely the scalable extension of the HEVC,is proposed.Apart from the high efficiency,the advanced technologies inherited from HEVC result in the significantly increased encoding complexity to SHVC.In order to effectively reduce the encoding time,two fast intra-frame prediction algorithms and a fast inter-frame prediction algorithm are proposed for SHVC in this thesis,which will make the real-time application of the SHVC possible.The main contents of the thesis are as follows:The proposed fast CU coding depth level decision algorithm for the enhancement layer in the spatially scalable HEVC intra-coding is based on the AC energy,which is used to classify the texture complexity of a CTU.For CTUs containing simple or rich texture,evaluations for some unlikely depth levels can be avoided.If the current CTU in the enhancement layer is located in the area with general texture,the inter-layer correlation is ultilised to determine a narrowed coding depth level range.Consequently,the coding depth level of the spatial adjacent CTUs of the current CTU are employed to further restrict the coding depth level range.In comparision with the original SHM algorithm,the encoding time is reduced by up to 52%,while the bitrate increment is only 0.89% and the decrease of BDPSNR is negligible(0.03 d B).The proposed fast coding mode decision algorithm for the enhancement layer in the spatially scalable HEVC intra-coding makes full use of inter-layer,temporal and spatial correlation to construct individual initial candidate list for PUs with different sizes,so that the number of intra modes in the RMD process can be reduced.Besides,the number of candidate modes is further reduced by comparing the Hadamard cost and the modes with higher Hadamard cost are removed.Compared with the original SHM algorithm,the proposed fast intra mode decision algorithm achieves about 19.54% time reduction on average at an average cost of 0.06 d B decrease in PSNR or a 0.98% increase in bitrate.In addition,the proposed fast CU depth level decision algorithm is combined with the fast coding mode decision algorithm.Experimental results indicate that the integrated algorithm achieves 40% reduction in encoding time while the decrease in BDPSNR is 0.08 d B and the average increase in BDBR is no more than 1.6%,which is negligible.Finally,a fast PU partition mode decision algorithm for the enhancement layer in the spatially scalable HEVC inter-coding is proposed.The inter-layer correlation is ultilised to predict the probable partition modes for the current CU,thus the number of modes which require RDC calculations is decreased.If the partition modes of the co-located CU in the baes layer is 2N×2N,the motion homogeneity,coding depth level are combined to avoid unnecessary computation while maintaining the similar coding efficiency.Experimental results show that compared with the SHM original inter-frame algorithm,when the Lowdelay configuration is employed,the algorithm reduces the encoding time at 36.3% on average while the decrease in BDPSNR is lower than 0.1d B and the increase in BDBR is negligible.