应用于VVC无损帧内编码的残差二次预测算法

Residual Secondary Prediction Algorithm for VVC Lossless Intra Coding

在VVC无损帧内编码中,为了无失真地压缩视频,会跳过变换、量化等操作。未经过变换的帧内预测残差之间,存在较强的空间相关性。基于此,提出了一种残差二次预测(Residual Secondary Prediction,RSP)算法,来进一步降低残差能量。算法对经过帧内预测生成的残差样本进行二次预测,得到二次残差,选择绝对值之和较低的残差组作为最终的预测残差。在VVC测试模型(VTM)12.3上应用RSP算法,相较于基础的无损帧内编码,平均可以节省5.65%的比特率,编码时间增加14.6%,解码时间下降17.5%。RSP以增加较低编码复杂度为代价,高效地提升了VVC无损帧内编码的性能。

This paper proposes a residual secondary prediction(RSP)algorithm to further reduce residual energy.The algorithm performs a secondary prediction on the residual samples generated through intra prediction to obtain the secondary residual,and selects the residual group with lower absolute sum as the final prediction residual.Applying the RSP algorithm on the VVC test model(VTM)12.3 can save 5.65%bit-rate on average,increase encoding time by 14.6%,and decrease decoding time by 17.5%compared to the basic lossless intra coding.RSP efficiently improves the performance of VVC lossless intra coding at the cost of increasing low coding complexity.