基于失真传递的时域自适应量化算法
发布时间：2018-11-03 15:10
【摘要】：码率控制是视频编码器中的关键模块,其算法直接决定编码器率失真性能.视频编码帧间预测导致的编码失真会在时域产生传递效应,考虑该传递效应是优化码率控制算法性能的关键.宏块树码率控制是一种典型的时域量化控制算法,核心是根据编码单元失真传递量(相对传递代价ρ)自适应地调整量化参数(偏移量δ),合适的δ-ρ映射关系是宏块树量化控制算法的核心.宏块树算法采用基于经验的δ-ρ模型,对不同视频序列的普适性有待改进,模型准确度和精度也需进一步优化.针对上述问题,将竞争决策方法用于探索最优δ-ρ映射关系,提出了一种率失真性能优化的失真时域传递自适应量化δ-ρ模型,以改进时域自适应量化算法.实验结果表明,信噪比BD-PSNR较原模型提升了0.14dB以上,SSIM性能提升了0.29dB.算法能更好地控制码率时域分配,降低失真时域传递恶化.
[Abstract]:Rate control is a key module in video encoder, and its algorithm directly determines the rate distortion performance of encoder. The coding distortion caused by inter-frame prediction of video coding will produce transfer effect in time domain, which is the key to optimize the performance of the rate control algorithm. Macroblock rate control is a typical time-domain quantization control algorithm, which adaptively adjusts quantization parameters (offset 未) according to the distortion transfer amount (relative transfer cost 蟻) of the coding unit. The appropriate 未-蟻 mapping relation is the core of the macro block tree quantization control algorithm. The 未-蟻 model based on experience is adopted in macroblock tree algorithm. The universality of different video sequences needs to be improved, and the accuracy and accuracy of the model need to be further optimized. To solve the above problems, the competitive decision method is applied to explore the optimal 未-蟻 mapping relationship, and a time domain transfer adaptive quantization 未-蟻 model is proposed to improve the time domain adaptive quantization algorithm. The experimental results show that the signal-to-noise ratio (SNR) of BD-PSNR is higher than that of the original model, and the performance of SSIM is improved by 0.29 dB. The algorithm can better control rate time domain allocation and reduce distortion time domain transmission deterioration.
【作者单位】： 中国计量大学信息工程学院;
【基金】：国家自然科学基金资助项目(61572449) 浙江省自然科学基金资助项目(LY15F020022,LY13H180011)
【分类号】：TN919.81
[Abstract]:Rate control is a key module in video encoder, and its algorithm directly determines the rate distortion performance of encoder. The coding distortion caused by inter-frame prediction of video coding will produce transfer effect in time domain, which is the key to optimize the performance of the rate control algorithm. Macroblock rate control is a typical time-domain quantization control algorithm, which adaptively adjusts quantization parameters (offset 未) according to the distortion transfer amount (relative transfer cost 蟻) of the coding unit. The appropriate 未-蟻 mapping relation is the core of the macro block tree quantization control algorithm. The 未-蟻 model based on experience is adopted in macroblock tree algorithm. The universality of different video sequences needs to be improved, and the accuracy and accuracy of the model need to be further optimized. To solve the above problems, the competitive decision method is applied to explore the optimal 未-蟻 mapping relationship, and a time domain transfer adaptive quantization 未-蟻 model is proposed to improve the time domain adaptive quantization algorithm. The experimental results show that the signal-to-noise ratio (SNR) of BD-PSNR is higher than that of the original model, and the performance of SSIM is improved by 0.29 dB. The algorithm can better control rate time domain allocation and reduce distortion time domain transmission deterioration.
【作者单位】： 中国计量大学信息工程学院;
【基金】：国家自然科学基金资助项目(61572449) 浙江省自然科学基金资助项目(LY15F020022,LY13H180011)
【分类号】：TN919.81
【相似文献】
相关期刊论文 前10条
1 王红玉;管鑫;周洪敏;;基于场景切换的码率控制方法[J];科技资讯;2008年03期
2 朱仲杰;白永强;蒋刚毅;粱丰;;基于双模预测的精确码率控制[J];中国图象图形学报;2010年04期
3 董焱鑫,李桂苓;模块化联合码率控制技术[J];电子技术应用;2000年08期
4 范晓东,余松煜,孙军;基于场景切换检测的码率控制策略[J];上海交通大学学报;2000年06期
5 环，
本文编号：2308154
本文链接：https://www.wllwen.com/kejilunwen/xinxigongchenglunwen/2308154.html
