分布式视频压缩感知中基于边信息及其支撑集的稀疏重构

文中提出了一种基于边信息及其支撑集的分布式视频压缩感知重构方案,它是建立在压缩感知和分布式视频编码理论的基础上,具有终端简单、采样/编码复杂度极低的特点.在该重构方案中,首先利用边信息构造当前（非关键）帧的冗余字典,为压缩感知重构提供稀疏表示方法;然后利用边信息估计的信号像素值和支撑集,辅助当前帧稀疏重构,提出了一种基于交替方向乘子的迭代算法以实现该重构优化问题的求解.实验结果表明,该重构方案能够获得较好的采样率和失真性能,主观图像质量也有所提高.

Sparse Reconstruction with Side Information and Its Support for Distributed Compressed Video Sensing

A novel sparse reconstruction scheme with side information（SI） and its support is presented for distributed compressed video sensing（ DCVS）, which integrates both compressed sensing and distributed video coding characteristics to achieve a low-complexity video coding. In the proposed scheme, a redundant dictionary is firstly built for the current non-key frame. Then, the estimated signal support and values from SI are used to perform an adaptive sparse reconstruction. Moreover, an alternative direction method of multipliers （ADMM） based numerical algorithm is presented to solve the underlying optimization problem. Simulation results show that the scheme can achieve an improved sampling rate distortion performance with better subjective image quality.