保留边缘与细节的压缩采样视频复原算法*

已有的压缩感知视频复原算法因过平滑效应难以保留视频帧的边缘与细节信息,对此提出一种基于混合稀疏性测量的压缩采样视频复原算法。编码端将视频序列分为关键帧与非关键帧,并使用相同的感知矩阵对帧的每块进行采样。解码端则设计了考虑局部稀疏性与全局稀疏性的混合稀疏性测量方案,并将其作为压缩感知视频复原问题的正则项；然后,通过分裂Bregman迭代算法对关键帧进行解码,并考虑视频帧间的时间相关性对非关键帧进行细化处理。基于多组仿真实验的结果表明,本算法获得了较好的视频复原精度,并具有理想的计算时间性能。

Compressive sampling video restore algorithm with edge and detail preserving

The existing restore algorithms of compressive sensing video are difficult to preserve the side and detail information of video frames due to the over-smoothing effect, a hybrid sparsity measurement based restore algorithm for compressive sampling video is proposed to solve that problem. In the encoding phase, video sequence is divided into key frames and non-key frames, and the same sensing matrix is used to sample each patches of the frame. In the decoding phase, a hybrid sparsity measurement schema considering local sparsity and global sparsity is designed, and the regularization term is replaced by the proposed hybrid sparsity measurement; then, the key frames are decoded by split Bregman iteration algorithm, and the non-key frames are refined by considering video inter-frame temporal correlation. Several simulation experimental results show that the proposed algorithm realizes better video restore accuracy, at the same time it shows a ideal computational complexity performance.