基于感知的多视点视频编码宏块模式选择快速算法

多视点视频编码(MVC)采取可变块模式选择技术和 多参考帧技术显著提高了编码的压缩效率,但同时带来了巨大的 编码计算复杂度。为了降低MVC的计算复杂度,提出基于感知的快速MVC宏块模式选择 算法。基于人眼视觉感知的特点,利用视觉恰可察觉失真(JND)的概念建立MVC宏块的最优模 式和JND的联系,并利用该联系确定早期结束最优宏块模式选择过程的阈值,根据当前编码 宏块的JND与阈值的 关系自适应地减少每个编码宏块的模式搜索次数,进而减少MVC的方向搜索和参考帧搜索的 次数,以降 低编码的复杂度,提高MVC速度。实验结果显示,对于不同运动特性、内容、纹理信息、相 机间距和图像 尺寸的测试序列,提出的快速算法在率失真性能几乎不变的情况下平均节约76.00% 编码时间。

A fast macroblock mode selection algorithm for multiview video coding based on visual perception

Multiview video coding (MVC) adopts ex haustive variable size mode decision and multiple reference frames selection to achieve s ignificantly high compression efficiency at each macroblock.However,the technology increases the computational complexity of MVC encoders tremendously.To solve this problem,a n ew fast macroblock mode selection algorithm of MVC based on visual perception is propose d by utilizing the conception of just noticeable difference (JND).It builds the correlation of MVC macroblock mode se lection with JND considering human visual perceptual properties to gain the threshold to optimize the process of va riable size mode selection in MVC.According to the relation between the threshold and the JND of the current macroblock,th e proposed algorithm can adaptively decrease the time of macroblock partition modes selecting,and then reduce the time of s earching for the best reference frame index and the best reference direction for each macroblock.That is to say,it can decrease co mplexity of each macroblock to improve the efficiency.Experimental results show that the proposed mode selection algorithm can reduce 76.00% of the encoding time averagely compared with JMVC and almost maintain the same coding e fficiency, while showing the adaptability to various video contents,camera arrangements and motion properties of scenes.