欢迎您！东篱公司
退出
-
申报数据库
-
立项数据库
-
成果数据库
-
项目获奖数据库
中文摘要：
率失真优化（RDO）是视频编码压缩中的关键技术之一。传统的率失真优化技术使用误差平方和（SSE）或绝对差和（SAD）度量失真,不能获得较好的视觉感知质量。本文针对新一代视频编码标准高效视频编码（HEVC）帧内编码RDO,提出了一种基于结构相似度（SSIM）度量失真的RDO方法。首先,根据HEVC编码树形单元（CTU）结构设计了基于SSIM的失真计算方法;然后,提出了一种基于量化器推导的码率-量化步长（R-Δ）模型和一种基于统计分析的失真-量化参数（DSSIM-QP）模型,用于求取拉格朗日乘数;最终,使用多QP优化方法求取模型参数。实验表明,相对于HEVC传统的RDO,针对全I帧（即所有帧都为帧内编码）固定QP编码和多QP优化编码,在相同的SSIM条件下,码率分别平均降低8.4%和13.9%左右,同时编码复杂度分别平均增加约3%和2%。
英文摘要：
Rate distortion optimization （REK）） is one of the key techniques in the field of video coding. Either sum of square error （SSE） or sum of absolute difference （SAD） is widely adopted to evaluate distortion in conventional RDO, which fails to achieve high consistency with human visual perception. Re cently, the structural similarity （SSIM） index has been attracting more and more attention as a perceptual visual quality metric for optimizing video codecs. Meanwhile, high efficiency video coding （HEVC） as the emerging video coding standard creates many new opportunities and challenges. In this paper, we propose a method of structural similarity index based RDO for HEVC intra coding. Firstly, an SSIM based distortion computational method is designed according to the coding tree unit （CTU） structure of HEVC. Secondly,a quantizer-derived rate-quantization step （R-A） model and a statistical analysis-based distortion-quantization parameter （DsslM QP） model are presented for Lagrange multiplier calculation. Finally,the model parameters are estimated based on multi QP optimization. Extensive experimental re sults show that compared with traditional RDO approach of HEVC,for the same SSIM score,the pro posed approach can achieve an average rate reduction of about 8.4% and 13.9% with 3% and 2% encoding time increasement, respectively for fixed QP and multi-QP optimizations on all intra coding configure.
同期刊论文项目
同项目期刊论文
Asymmetric self？recovery oriented stereo image watermarking method for three dimensional video syste
New Stereoscopic Image watermarking Method Based on Correlation of Inter-view Block Pair for 3D Medi
期刊信息
位置：