The new features of H. 264 video coding standard make the motion estimation module much more time consuming than before. Especially, the motion search is required for each of the 4 modes for inter prediction. In order...The new features of H. 264 video coding standard make the motion estimation module much more time consuming than before. Especially, the motion search is required for each of the 4 modes for inter prediction. In order to reduce the computational complexity, we analyze the statistics of results of motion estimation, such as the continuity of best modes of blocks in successive frames and the chance to give up a sub-partition mode （smaller than 16 × 16） after integer-pixel motion estimation, from which we suggest to make mode prediction based on the motion information of the previous frame and skip sub-pixel motion estimation in subpartition mode selectively. According to the experimental result, the proposed algorithm can save 75 % of the computational time with a slight degradation （0.03 dB） on PSNR compared with the pseudocode of fast search motion estimation in JM12.2.展开更多
针对高效视频编码(HEVC)帧内预测过程中的高计算复杂度问题,提出一种基于纹理特征的预测模式选择和编码单元划分的快速帧内预测算法。利用每一深度层纹理方向强度判断编码单元是否需要进行分割,并且减少候选模式数量。首先,在每一深度...针对高效视频编码(HEVC)帧内预测过程中的高计算复杂度问题,提出一种基于纹理特征的预测模式选择和编码单元划分的快速帧内预测算法。利用每一深度层纹理方向强度判断编码单元是否需要进行分割,并且减少候选模式数量。首先,在每一深度层编码单元上结合像素方差,以像素点为单位计算相应的纹理方向强度,确定其纹理复杂度并结合阈值策略预测最终划分深度;其次,比较垂直和水平方向强度关系及统计预测候选模式概率分布,以减少预测模式数量,确定最优候选模式子集,进一步降低编码复杂度。所提算法与平台HM15.0相比,编码时间平均节省51.997%,BDPSNR(Bjontegaard Delta Peak Signal-to-Noise Rate)仅降低0.059 d B,BDBR(Bjontegaard Delta Bit Rate)仅上升了1.018%。实验数据表明,在保证信噪比和比特率基本不变的同时,所提算法能有效降低编码复杂度,利于HEVC的实时视频应用。展开更多
针对高效率视频编码(high efficiency video coding,HEVC)标准编码复杂度较高的问题,提出了一种快速编码单元(coding unit,CU)划分方法。首先,结合拉格朗日率失真优化理论及相关实验数据分析得到相邻CU深度对应的失真及码率分别满足线...针对高效率视频编码(high efficiency video coding,HEVC)标准编码复杂度较高的问题,提出了一种快速编码单元(coding unit,CU)划分方法。首先,结合拉格朗日率失真优化理论及相关实验数据分析得到相邻CU深度对应的失真及码率分别满足线性关系,并利用此关系,建立了率失真代价的预测模型。利用此模型,可以在编码当前CU深度后快速预测得到下一CU深度的率失真代价,并最终通过代价比较,判断是否需要继续进行CU划分。实验结果表明,相比于HEVC测试模型HM12.0,针对低时延与随机接入编码结构,提出的方法的BD-rate分别增加了0.2%与0.6%,同时,编码时间分别减少了33.2%和38.9%。展开更多
基金Sponsored by the National Natural Science Foundation of China(60772066)
文摘The new features of H. 264 video coding standard make the motion estimation module much more time consuming than before. Especially, the motion search is required for each of the 4 modes for inter prediction. In order to reduce the computational complexity, we analyze the statistics of results of motion estimation, such as the continuity of best modes of blocks in successive frames and the chance to give up a sub-partition mode （smaller than 16 × 16） after integer-pixel motion estimation, from which we suggest to make mode prediction based on the motion information of the previous frame and skip sub-pixel motion estimation in subpartition mode selectively. According to the experimental result, the proposed algorithm can save 75 % of the computational time with a slight degradation （0.03 dB） on PSNR compared with the pseudocode of fast search motion estimation in JM12.2.
文摘针对高效视频编码(HEVC)帧内预测过程中的高计算复杂度问题,提出一种基于纹理特征的预测模式选择和编码单元划分的快速帧内预测算法。利用每一深度层纹理方向强度判断编码单元是否需要进行分割,并且减少候选模式数量。首先,在每一深度层编码单元上结合像素方差,以像素点为单位计算相应的纹理方向强度,确定其纹理复杂度并结合阈值策略预测最终划分深度;其次,比较垂直和水平方向强度关系及统计预测候选模式概率分布,以减少预测模式数量,确定最优候选模式子集,进一步降低编码复杂度。所提算法与平台HM15.0相比,编码时间平均节省51.997%,BDPSNR(Bjontegaard Delta Peak Signal-to-Noise Rate)仅降低0.059 d B,BDBR(Bjontegaard Delta Bit Rate)仅上升了1.018%。实验数据表明,在保证信噪比和比特率基本不变的同时,所提算法能有效降低编码复杂度,利于HEVC的实时视频应用。
文摘针对高效率视频编码(high efficiency video coding,HEVC)标准编码复杂度较高的问题,提出了一种快速编码单元(coding unit,CU)划分方法。首先,结合拉格朗日率失真优化理论及相关实验数据分析得到相邻CU深度对应的失真及码率分别满足线性关系,并利用此关系,建立了率失真代价的预测模型。利用此模型,可以在编码当前CU深度后快速预测得到下一CU深度的率失真代价,并最终通过代价比较,判断是否需要继续进行CU划分。实验结果表明,相比于HEVC测试模型HM12.0,针对低时延与随机接入编码结构,提出的方法的BD-rate分别增加了0.2%与0.6%,同时,编码时间分别减少了33.2%和38.9%。
