一种基于零块分布的快速帧间模式判决算法
发布时间：2018-10-30 11:59
【摘要】：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。
[Abstract]:Aiming at the high computational complexity of the new generation of discrete visual coding (HEVC) inter-frame predictive coding, a low complexity algorithm for fast inter-frame mode filtering using the transformed quantized zero-block distribution feature is proposed. Firstly, the regional distribution characteristics of zero coefficient and non-zero coefficient under SKIP and inter-frame 2N 脳 2 N mode are statistically analyzed, and the motion characteristics and motion estimation matching effect are judged to analyze the distribution of SKIP mode under zero block. At the same time, the PU pattern distribution is selected by combining the SATD cost classification of the residual block with the non-zero block, and the PU mode which accords with the zero-block distribution feature is selected first, so that the inter-frame mode can be filtered quickly and the computational complexity can be reduced. The experimental results show that under the condition of low delay and random access configuration, the coding time of the proposed algorithm can be reduced by 50.3 and 49.3, respectively, compared with the original HM fast algorithm. BDBR only reduces the encoding complexity by 0.03%.
【作者单位】： 上海大学通信与信息工程学院;上海市消防总队浦东支队;
【基金】：国家自然科学基金(61271212) “863”计划(2015AA015903) 上海市自然基金项目(14ZR1415200) 上海市“科技创新行动计划”(14511105602)资助项目
【分类号】：TN919.81
[Abstract]:Aiming at the high computational complexity of the new generation of discrete visual coding (HEVC) inter-frame predictive coding, a low complexity algorithm for fast inter-frame mode filtering using the transformed quantized zero-block distribution feature is proposed. Firstly, the regional distribution characteristics of zero coefficient and non-zero coefficient under SKIP and inter-frame 2N 脳 2 N mode are statistically analyzed, and the motion characteristics and motion estimation matching effect are judged to analyze the distribution of SKIP mode under zero block. At the same time, the PU pattern distribution is selected by combining the SATD cost classification of the residual block with the non-zero block, and the PU mode which accords with the zero-block distribution feature is selected first, so that the inter-frame mode can be filtered quickly and the computational complexity can be reduced. The experimental results show that under the condition of low delay and random access configuration, the coding time of the proposed algorithm can be reduced by 50.3 and 49.3, respectively, compared with the original HM fast algorithm. BDBR only reduces the encoding complexity by 0.03%.
【作者单位】： 上海大学通信与信息工程学院;上海市消防总队浦东支队;
【基金】：国家自然科学基金(61271212) “863”计划(2015AA015903) 上海市自然基金项目(14ZR1415200) 上海市“科技创新行动计划”(14511105602)资助项目
【分类号】：TN919.81
【相似文献】
相关期刊论文 前10条
1 杨越;高新波;冯s，
本文编号：2299948
本文链接：https://www.wllwen.com/kejilunwen/xinxigongchenglunwen/2299948.html
