一种基于视觉显著性的码率控制算法

A Bit Rate Control Algorithm Based on Visual Saliency

针对当前高效视频编码的码率控制算法未能充分确保视频显著区域的图像质量问题,提出了一种基于视觉显著性的码率控制算法。首先,采用基于图的流形排序显著检测算法获取视频的显著区域;其次,在图像编码过程中,为每个编码单元(Coding Unit,CU)建立基于视觉显著性的加权失真公式,将显著性值作为权重;最后,通过求解得到拉格朗日乘子λ,并根据原始码率控制算法中的λ与量化参数(Quantization Parameter,QP)的关系为每个CU确定QP。试验结果表明,该算法有效地根据视频内容的显著性权重实现了码率的合理分配,并在保持码率控制准确性的同时,提升了视频显著区域的图像质量。

To address the problem that current bit rate control algorithms for efficient video coding fail to adequately ensure the image quality of video salient regions,this paper proposes a visual saliency-based bit rate control algorithm.First,a graph-based popular ordering saliency detection algorithm is used to obtain the salient regions of the video.Then,a weighted distortion formula based on visual saliency is established for each CU(Coding Unit)in the image coding process,and the saliency values are used as weights.Finally,the Lagrange multiplierλis obtained by solving,and the QP(Quantization Parameter)is determined for each CU based on the relationship between theλand the QP in the original bit rate control algorithm.The results indicate that the algorithm effectively achieves the reasonable allocation of the bit rate according to the saliency weights of the video content,and improves the image quality of the video salient regions while maintaining the accuracy of the bit rate control.