基于感知分析的立体视频帧重要性模型

A Frame Importance Model of Stereoscopic Video Based on Perception Analysis

由于丢失不同帧对立体视频质量的终端感知影响不同,因此提出一个面向内容感知的帧重要性区分模型.根据立体视频序列的时空相关性及视频帧之间错误扩散的特性,新模型从帧丢失引起的当前帧的错误隐藏失真和帧丢失引起后续帧的错误扩散失真两方面来计算帧重要性.实验结果表明:新模型能够在编码端准确地估算各帧对终端感知的重要性,对不同运动剧烈程度和视差大小的立体视频序列都具有适用性.

The loss of frames affects the perception quality of stereoscopic video at terminal, to tackle this problem, a model of frame importance is proposed on the basis of content of stereo video. According to the spatial and temporal correlation, and the characteristics of inter-frame error diffusion, the model calculates the frame importance from the error concealment distortion and error diffusion distortion caused by the loss of current frame. The results show that the model can estimate the importance of each frame satisfactorily, and can be adapted to video sequences with different motion intensity and disparity.