TurboZoom (Zoomer) performs image super-resolution by solving a regularized least-squares problem using an efficient conjugate gradient method. The algorithm reconstructs high-resolution images from low-resolution blurred measurements by inverting the forward model that combines Gaussian blur and binning (downsampling). It includes zeroth-order (ridge) and first-order (smoothness) Tikhonov regularization for stable reconstruction. The method is particularly effective for microscopy, satellite imagery, and hyperspectral imaging where spatial resolution is limited by sensor binning or optical blur. No training data or neural networks required: works directly from single images.
Paul H. C. Eilers [1]Cyril Ruckebusch [2]
[1]Department of Biostatistics, Erasmus Medical Center, Rotterdam, The Netherlands.[2]LASIRE CNRS, University of Lille, Lille, France.
