Frequency-Domain Image Processing
Overview
Frequency-domain image processing studies how images’ spatial patterns can be represented as frequencies and used to recover, enhance, segment, or combine visual information. Common approaches use Fourier transforms or wavelets, separate low- and high-frequency features, model phase as well as amplitude, and combine spectral processing with spatial networks through attention, unrolled optimization, or diffusion guidance. Applications include deblurring and super-resolution on benchmarks such as GoPro and RealBlur, low-light enhancement, medical segmentation, and infrared-visible image fusion. A central challenge is balancing global spectral context and efficient computation with precise local detail, while ensuring methods generalize across different degradations, sensors, and datasets.
Latest papers
30 papers5 papers in the last four weeks, up 67% on the four weeks before. 0.0% of all new papers.