针对新一代视频编码标准离散视编码(HEVC)帧间预测编码计算复杂度过高的问题,提出一种利用变换量化后零块分布特征进行快速帧间模式筛选的低复杂度算法。首先统计SKIP和帧间2 N×2 N模式下零系数与非零系数的区域分布特征,判断其运动特征和运动估计匹配效果,用以分析全零块下SKIP模式的分布;同时在非全零块下结合残差块的SATD代价分类筛选PU模式分布,从而优先选择符合零块分布特征的PU模式,以达到快速筛选帧间模式,降低计算复杂度的效果。实验结果表明,在低延迟和随机访问配置条件下,本文提出的算法在保证编码性能的情况下,相比原有HM快速算法,编码时间平均分别可减少50.3和49.3%,BDBR仅降低0.03%,有效地降低了编码复杂度。
Aiming at the higher computational complexity caused in inter predictive coding for a new generation of high efficiency video coding (HEVC), a low complexity coding algorithm guided by fast intermode decision using the zero-block distribution after transformation and quantization is proposed in the paper. First, the regionally distributive characteristics of zero coefficients and non-zero coefficients are recorded,which is used to analyze the distributions of SKIP mode in all zero block and the prediction unit (PU) mode uniting the cost of sum of absolute transformed difference (SATD) in not all zero block,on the basis of the match effect of motion characteristics and motion estimation. So the PU mode conformed to the distribution of zero block is selected preferentially for filtering inter modes and reducing computational complexity. On the condition of random access and low-delay configuration,the experiment results indicates that the encoding performance is guaranteed by the proposed algorithm in this paper. Mean- while comparing with the original algorithm of HM,encoding time is also further reduced by 50% on average. So the coding computational complexity is dropped greatly in the paper.