2020/03/16 by Majed El Helou, Helou, Majed El, Ruofan Zhou +3 · 1 citation
Computer Science · Engineering · #Advanced Image Processing Techniques #Computer Vision and Pattern Recognition (cs.CV) #FOS: Computer and information sciences #FOS: Electrical engineering #Image Processing Techniques and Applications #Image and Video Processing (eess.IV) #Photoacoustic and Ultrasonic Imaging #electronic engineering #information engineering
paper · pdf · doi:10.48550/arxiv.2003.07119
openalex publication_date 2020/03/16 · openalex created_date 2025/10/10 · openalex updated_date 2026/07/28
Super-resolution and denoising are ill-posed yet fundamental image\nrestoration tasks. In blind settings, the degradation kernel or the noise level\nare unknown. This makes restoration even more challenging, notably for\nlearning-based methods, as they tend to overfit to the degradation seen during\ntraining. We present an analysis, in the frequency domain, of\ndegradation-kernel overfitting in super-resolution and introduce a conditional\nlearning perspective that extends to both super-resolution and denoising.\nBuilding on our formulation, we propose a stochastic frequency masking of\nimages used in training to regularize the networks and address the overfitting\nproblem. Our technique improves state-of-the-art methods on blind\nsuper-resolution with different synthetic kernels, real super-resolution, blind\nGaussian denoising, and real-image denoising.\n