探讨了提高H.264视频编码性能的场景自适应比特率控制问题.通过计算高维空间中多模式聚集的最大特征向量距离,提出了基于巴氏距离的场景变化度量,并有效采用低秩Cholesky分解计算近似巴氏距离.在此基础上,提出了自适应的GOP结构和量化参数调整策略.实验模拟了视频样本序列.由于具有差错复原和较低的计算复杂度等特点,实验结果表明该方法十分有效.%This paper studies scene adaptive bit rate control to improve the performance of existing H.264 video codec. By maximizing the feature vector distance between multi-modal clusters in a hyper-sphere space, a Bhattacharyya distance based metrics has been utilized to quantify the scene change. Moreover, a low rank Cholesky factorization has been adopted to compute the approximate Bhattacharyya distance quite efficiently. Then an adaptive GOP structure and quantization parameter adjustment is proposed to optimize the rate control. Video sequence in the video trace database is simulated and the experimental results demonstrate that our proposed approaches are quite effective with the characteristics of the error resilience to network packet losses and low complexity.
展开▼
