A Lightweight Recurrent Learning Network for Sustainable Compressed Sensing

Recently, deep learning-based compressed sensing (CS) has achieved great success in reducing the sampling and computational cost of systems improving reconstruction quality. These approaches, however, largely overlook issue cost; they rely on complex structures task-specific operator designs, resulting extensive storage high energy consumption CS imaging systems. In this article, we propose a lightweight but effective neural network based recurrent learning to achieve sustainable system; it requires smaller number parameters obtains high-quality reconstructions. Specifically, our proposed consists an initial sub-network residual sub-network. While hierarchical structure progressively recover image, parameters, facilitates feature extraction via perform both fusion reconstructions across different scales. addition, also demonstrate that, after reconstruction, maps with reduced sizes are sufficient information, thus significant reduction amount memory required. Extensive experiments illustrate that model can better quality than existing state-of-the-art algorithms, these algorithms. Our source codes available at: https://github.com/C66YU/CSRN .