Study Of Fast Mode Decision In HEVC Transcoding Algorithm And Parallel Implementation Of Deblock Filter

The new generation video coding standard HEVC can achieve better performance than H.264 in the range of 50% bitrate reduction for equal perceptual video quality. Therefore, HEVC is becoming the main standard of the industry. HEVC Ultra HD encoding is one of the research hotspots. In addition, there is a large amount of content encoded in H.264 to be converted to HEVC format. Fast H.264 to HEVC transcoding algorithm is in great need. This paper includes the following two aspects: parallel implementation of deblocking filter in HEVC encoder, and fast mode decision in transcoding HEVC algorithm.Deblocking filter can effectively reduce the artifacts of the encoder and improve the performance significantly. Based on the analysis of the dependency of the deblocking filter, we proposed a feasible CTU-level parallelism scheme. Meanwhile, we designed the data storage structure and implemented deblocking filter on the multi-core platform with low complexity.For HEVC transcoding, this paper first presents a fast mode decision transcoding algorithm based on residual homogeneity. It mainly focuses on coding unit(CU) and prediction unit(PU) decision by using blocks’ homogeneity characteristic. The residual homogeneity indicator(RHI) and sum of absolute residual(SAR) value are extracted to early terminate CU partition. Further, the unnecessary PU partitions are excluded by exploring RHI information from H.264/AVC bitstream. Experimental results illustrate that the proposed algorithm achieves about 56.68% complexity saving in terms of encoding time with only 2.3% average bitrate increase when compared to the trivial transcoder. In addition, a new transcoding algorithm based on prediction homogeneity is proposed. The motion homogeneity indicator(MHI) is added to improve the accuracy of the fast PU mode decision. And the coded block flag(CBF) imformation of 2Nx2 N partition is also used to make decision for early CU split termination. The experimental results show that the proposed transcoding scheme can save up to 50% of encoding time with 2.08% average bitrate increase when compared to the trivial transcoder.