激光与光电子学进展, 2017, 54 (7): 071003, 网络出版: 2017-07-05
基于图像纹理特征的HEVC帧内编码快速算法 下载: 941次
Fast Intra-Frame Encoding Algorithm Based on Image Texture Features for HEVC
图像处理 高效视频编码 纹理复杂度 编码单元划分 模式选择 image processing high efficiency video coding texture complexity coding unit partitioning mode decision
摘要
为降低高效视频编码(HEVC)的帧内编码复杂度,提出一种基于图像纹理特征的编码单元(CU)划分和预测模式选择算法。利用一种预处理算法来获得当前CU的纹理复杂度和方向。一方面,根据CU的纹理复杂度,该算法自适应地跳过或终止部分CU划分,减少CU深度的遍历时间。纹理复杂度高的CU直接划分成4个子CU,纹理复杂度低的CU将会终止划分。另一方面,根据预测单元的纹理方向,确定相应的候选模式集,通过粗模式决策算法和率失真最优化算法遍历候选模式集选取最优模式。将算法移植到标准食品解码软件HM16.7平台上,实验结果表明,与HM16.7算法相比,编码时间平均减少53.66%,比特率上升0.46%,峰值信噪比下降0.05 dB。
Abstract
In order to reduce the intra encoding complexity of the high efficiency video coding (HEVC), a novel algorithm is proposed, which is composed of the coding unit (CU) partitioning and the prediction mode decision based on image texture features. A preprocessing algorithm is employed to get the texture complexity and the direction of the current CU. On the one hand, according to the texture complexity of CU, the proposed algorithm can adaptively skip or suspend some CU divisions, and thus reduce the depth traversal time of the CU. The CU with high texture complexity is divided into four smaller equal-size sub-CUs directly. As for the CU with low texture complexity, the searching of partitioning is early terminated. On the other hand, a set of alternative patterns are confirmed according to the dominant texture direction of the prediction unit. the optimal model is chosen for the further rough mode decision algorithm and rate distortion optimization processes by traveling the alternative patterns. Moving the algorithm to the HM16.7, a normal decode software for food, the proposed algorithm can save 53.66% coding time on average, the negligible bit rate increases by 0.46%, and the peak signal-to-noise ratio reduces by 0.05 dB.
孙学斌, 陈晓冬, 肖禹泽, 汪毅, 郁道银. 基于图像纹理特征的HEVC帧内编码快速算法[J]. 激光与光电子学进展, 2017, 54(7): 071003. Sun Xuebin, Chen Xiaodong, Xiao Yuze, Wang Yi, Yu Daoyin. Fast Intra-Frame Encoding Algorithm Based on Image Texture Features for HEVC[J]. Laser & Optoelectronics Progress, 2017, 54(7): 071003.