基于多帧参考的整帧丢失错误掩盖算法

为了解决视频传输中整帧图像丢失的误差扩散问题,提出了一种基于多参考帧运动矢量外推的整帧丢失错误掩盖算法.通过对多个参考帧的运动矢量外推得到丢失块的候选运动矢量集|采用多参考帧边界匹配准则进行丢失帧和其后续帧的误差估计,并选择丢失块的最优运动矢量|最后采用自适应的重叠块运动补偿方法重建丢失帧.实验结果表明,该算法在恢复整帧图像的同时,有效地降低了视频序列中整帧图像丢失的误差扩散影响,与已有算法比较,该算法恢复的视频序列平均峰值信噪比提高了0.5~1 dB,具有更好的错误掩盖效果.

An Error Concealment Algorithm for Whole Frame Loss Based on Multiple-frame Extrapolation

To solve the error propagation of whole frame loss in video transmission,a whole frame error concealment algorithm based on multiple-frame extrapolation is presented.Bidirection motion vector extrapolation method from available neighboring frames is used to get the motion vection candidate sets of the lost frame.A weighted multi-frame boundary matching errors criterion is presented to estimate the distortion of both the lost frame and its succeeding frame,and the optimal motion vection is determined from the motion vection candidate sets.Finally,adaptive overlapped blockmotion compensation is used to reconstruct the lost frame.Simulation results show that the presented algorithm can recover the lost frame and effectively suppress error propagetion.Compared with other conventional methods,the proposed algorithm achieves 0.5~1 dB gain in terms of average PSNR and provides much better error concealment results.