Degradation-aware deep unfolding network with transformer prior for video compressive imaging

Jianfu Yin, Nan Wang, Binliang Hu, Yao Wang, Quan Wang

Abstract: Highlights•Introduces a novel approach for video snapshot compressive imaging reconstruction.•Utilizes estimated priors from compressed frames and physical masks to guide iterative learning.•Innovates by capturing both intra-frame contents and inter-frame dependencies simultaneously.•Integrates BiP-CRNN into DADUN to enhance the ability to handle long-range dependencies, forming a robust framework for video sequence reconstruction.•Incorporates the Transformer priors to further improve reconstruction quality.

External IDs:dblp:journals/sigpro/YinWHWW25