基于双向预测的变采样率分块压缩视频感知

摘要

At present,the existing block compression video sensing(CVS)scheme based on variable sampling rate only uses the forward prediction of video frames.In this way,firstly,the reference frame is measured,and then the sampling rate is adaptively allocated according to the correlation of blocks between the current frame and the reference frame.This method only uses the correlation between the current frame and the previous frame,but ignores the relevance between the current frame and the next frame.In order to solve this problem,a novel adaptive rate block compressive video sensing scheme with bidirectional prediction is proposed.In this scheme,reasonable sampling rate is allocated to the current frame and the video images are reconstructed by using bidirectional prediction and bidirectional motion estimation.The experimental results show that the proposed scheme can reduce the sampling rate properly and improve the quality of video reconstruction compared with the existing variable sampling rate scheme.It can obtain a peak signal-to-noise ratio gain of 1-3 dB and a structural similarity gain of 0.02-0.06.%目前现有的变采样率分块压缩视频感知方案仅使用了视频帧的前向预测,即先对参考帧进行测量,然后根据当前帧与参考帧各块之间相关性的大小自适应地分配采样率.这种方式只利用了当前帧与参考帧的帧间相关性,而忽视了自身与后面帧的相关性特点.针对这个问题,提出了一种基于双向预测的变采样率分块压缩视频感知方案,利用双向预测和双向运动估计为当前帧分配合理的采样率并重构视频图像.实验结果表明:相比较现有的变采样率方案,此方案可以适当降低帧的采样率并且提高视频的重构质量,获得1～3 dB的峰值信噪比增益和0.02～0.06的结构相似性增益.