Image Compressed Sensing

Image compressed sensing (ICS) aims to reconstruct high-quality images from significantly fewer measurements than traditional methods, leveraging the inherent sparsity or low-rank structure of image data. Current research heavily utilizes deep learning, employing architectures like deep unfolding networks, transformers, and convolutional neural networks (often incorporating multi-scale or non-local features) to learn efficient sampling strategies and powerful reconstruction models. These advancements offer improved reconstruction accuracy and computational efficiency compared to classical methods, with implications for various applications including medical imaging, remote sensing, and video compression. A key focus is on developing adaptive sampling techniques that tailor measurement acquisition to image content, further enhancing compression ratios.