Compressed Sensing

Compressed sensing studies how to reconstruct signals from undersampled measurements, making acquisition faster or less costly in settings such as MRI, CT, and single-pixel imaging. Methods range from sparse optimization and sensing-matrix design to deep unfolding, plug-and-play denoisers, and generative priors; some adapt measurements to a downstream task rather than image fidelity alone. Work spans theoretical recovery guarantees and practical systems for images, videos, and other structured data. A central tension is that learned priors can improve reconstructions with very few measurements but may be sensitive to operator or prior mismatch, while stronger guarantees do not always predict perceptual or task performance.