基于前向预测并行追踪的视频压缩传感方法

摘要

为了减小视频信号采集端的负担，提出一种基于压缩传感理论的视频压缩算法模型。该模型采用广义轮换矩阵获取视频帧的测量值，再结合视频压缩的运动补偿预测关键技术，消除视频帧间冗余。然后在并行贪婪追踪算法中引入前向预测技术作为重构算法，最终获取重构的压缩图像。实验对比了在不同算法下的重构图像和不同采样比下不同算法的重构PSNR值，然后利用所提方法得到视频序列的重构压缩图像。实验结果表明，该算法模型能较精确地获取压缩的视频图像，适合在视频压缩中应用。%In order to reduce the load at video signal collection end, we propose a video compression algorithm model which is based on com-pressed sensing theory.The model adopts generalised rotation matrix to acquire the measured values of video frames and then combines the key technology of motion compensation prediction of video compression to remove redundancies between video frames.Next, it introduces the forward prediction technology to parallel greedy tracking algorithm as the reconstruction algorithm, and finally obtains the reconstructed com-pression images.In the experiments we compare the reconstructed images with different algorithms and the reconstructed PSNR values with different algorithms in different sampling ratios, and use the presented method to obtain the reconstructed compressive images of video se-quence.Experimental results indicate that more accurate compressed video images can be obtained by the proposed algorithm model, and it is suitable for application in video compression.