The article introduces the MsDC-DEQ-Net, a neural network model designed for image reconstruction using compressive sensing (CS). CS is a technique that recovers sparse signals using fewer measurements than traditional methods. The MsDC-DEQ-Net model integrates the ISTA block with aggregated residual transformations and squeeze-and-excitation mechanisms. The model exhibits competitive performance compared to other network-based methods, with reduced storage requirements and improved reconstruction accuracy. It also benefits from multi-scale dilated convolutions, further enhancing its performance. This model finds applications in various fields including MRI, radar signal sampling, cryptosystems, snapshot imaging, and video sensing.
Publication date: 8 Jan 2024
Project Page: ?
Paper: https://arxiv.org/pdf/2401.02884
