1: \begin{abstract}

2: Snapshot compressive imaging (SCI) aims to capture the high-dimensional (usually 3D) images using a 2D sensor (detector) in a single snapshot.

3:  Though enjoying the advantages of low-bandwidth, low-power and low-cost, applying SCI to large-scale problems (HD or UHD videos) in our daily life is still challenging.

4:  The bottleneck lies in the reconstruction algorithms; they are either too slow (iterative optimization algorithms) or not flexible to the encoding process (deep learning based end-to-end networks).  

5:  In this paper, we develop fast and flexible algorithms for SCI based on the plug-and-play (PnP) framework.

6:  In addition to the widely used PnP-ADMM method, 

7:  we further propose the PnP-GAP (generalized alternating projection) algorithm with a lower computational workload and prove the convergence\footnote{We have found an error in the proof of the camera ready version of the CVPR paper in the CVF website.

8:  		Specifically, the lower bound in Eq. (25) is wrong. Following this, the proved global converge of PnP-GAP does not hold. In this new version, we show another convergence proof of PnP-GAP.} of PnP-GAP under the SCI hardware constraints. 

9:  By employing deep denoising priors, we first time show that PnP can recover 

10:  a UHD color video ($3840\times 1644\times 48$ with PNSR above 30dB) from a snapshot 2D measurement. 

11:  Extensive results on both simulation and real datasets verify the superiority of our proposed algorithm. The code is available at \url{https://github.com/liuyang12/PnP-SCI}

12: \end{abstract}

13: