MAR-Net: Multi-Attention Residual Network for Lensless Image Super-Resolution

Aiming at the problems of blurred edge contours and low overall clarity of lensless images in the process of super-resolution reconstruction, this paper proposes an image super-resolution network MAR-Net combining multiple attention mechanisms. The algorithm strengthens the key region feature extraction by introducing the RSCAB module and combines the multiple attention mechanism module and residual structure, which effectively improves the reconstruction effect of the lensless images. Experimental results show that MAR-Net can significantly improve the edge clarity of the image, eliminate noise and blur, and improve the overall visual quality of the image. The method in this paper performs well in edge information extraction and detail recovery of cellular images, providing strong support for the research and application of lens-free images.