论文部分内容阅读
该文提出了一种基于恒定SNR的单次扫描帧级比特率分配的视频质量平滑方案(CSNRBA)。采用之前所有编码帧的SNR的递减几何级数加权平均值作为当前帧的目标SNR,再根据国际上最近提出的闭合形式D-Q分析模型和线性码率模型,准确计算当前帧的比特预算。实验结果表明,该文提出的CSNRBA算法提供了比保持PSNR恒定算法和目前广泛采用的恒定比特率分配算法都要平滑得多的视频质量,并在物体清晰度、轮廓和自然色彩方面明显改善了传输视频的主观质量。
This paper presents a video quality smoothing scheme (CSNRBA) based on constant SNR for single-scan frame-level bitrate allocation. The weighted average of the decreasing geometrical series of SNRs of all the previous coded frames is used as the target SNR of the current frame. Based on the recently proposed closed-form D-Q analysis model and linear rate model, the bit budget of the current frame is accurately calculated. The experimental results show that the CSNRBA proposed in this paper provides a much smoother video quality than the PSNR invariant algorithm and the constant bit-rate allocation algorithm, which is widely used at present, and obviously improves the object definition, contour and natural color The subjective quality of the transmitted video.