基于立体视频传输失真模型预测参数研究

Prediction Parameters for Stereoscopic Video Transmission Distortion Model: Study

针对现有的基于递归计算的像素级立体视频传输失真模型没有完善的预测参数求取算法,提出了一种基于立体视频传输失真模型的自适应立体视频预测参数的求取算法.该算法根据立体视频时空域特性,确定未丢失的帧间编码宏块采取的运动估计方法,并采用时域或者视间视差补偿的方式确定端到端预测失真及隐藏失真,进而确定时域和视间预测参数.结果表明:该算法与传统的预测参数固定取值算法相比,各视点的失真误差都有所减小,且对各种特性的序列有较好的适应性.

Aiming at the existing distortion model without reliable parameters for pixel-level stereo video transmission, and using the recursive computation, an adaptive stereo video predictive parameterization algorithm is proposed. Firstly, according to the temporal and spatial features of stereo video it obtains motion estimation with no inter frame macro block loss, then calculates the end-to-end distortion and concealment distortion using temporal compensation or inter video compensation to determine both the temporal and spatial prediction parameters. The experimental results show that the distortion error of all views have decreased significantly, and compared with the traditional single view video algorithm which has the predictive parameters of the fixed value, the proposed algorithm can obtain better adaptability to sequence with large range of varying features.