基于纹理区域分割的SVC快速模式决策算法

Fast Mode Decision Algorithm for SVC Based on Texture Region Segmentation

为降低可分级视频编码(SVC)模式决策的计算量,提出一种快速模式决策算法。利用INTRA模式率失真代价及其在层间的相关性,对增强层的图像纹理区域进行分割,使用提前终止阈值在SKIP模式、BLPRED模式中选择最佳模式,以INTRA模式率失真代价计算宏块的纹理区域类型,根据该区域类型以及运动活跃性,动态调整INTER模式的估计过程。实验结果表明,该算法能以很小的峰值信噪比损失,使编码时间减少52.58%。

Mode decision algorithm in Scalable Video Coding（SVC） has high computational complexity.In this paper,a fast mode decision algorithm is proposed.In the proposed algorithm,INTRA mode rate-distortion cost and its interlayer correlation are utilized to segment the enhancement picture into different texture region,the early termination threshold is utilized to determine the SKIP and BLPRED mode in advance,then,the INTRA mode rate-distortion cost is utilized to calculate the texture region type for the macroblock in enhancement layer.According to the texture region type and motion activity of the macroblock,the INTER mode estimation is dynamically adjusted.Experimental results show that the proposed algorithm can reduce the total encoding time by up to 52.58% with very low Peak Signal to Noise Ratio（PSNR） loss.