Title:Deep Unfolding-Aided Parameter Tuning for Plug-and-Play Based Video Snapshot Compressive Imaging

Abstract:Snapshot compressive imaging (SCI) captures high-dimensional data efficiently by compressing it into two-dimensional observations and reconstructing high-dimensional data from two-dimensional observations with various algorithms. Plug-and-play (PnP) is a promising approach for the video SCI reconstruction because it can leverage both the observation model and denoising methods for videos. This paper proposes a deep unfolding-based method for tuning noise level parameters in PnP-based video SCI, which significantly affects the reconstruction accuracy. For the training of the parameters, we prepare training data from the densely annotated video segmentation (DAVIS) dataset, reparametrize the noise level parameters, and apply the checkpointing technique to reduce the required memory. Simulation results show that the trained noise level parameters significantly improve the reconstruction accuracy and exhibit a non-monotonic pattern, which is different from the assumptions in the conventional convergence analyses of PnP-based algorithms.