GLOW-FM: Efficient Generative Low-Light Image Enhancement with One-Step Latent Flow Matching

Researchers introduced GLOW-FM, a generative low-light image enhancement method that uses one-step latent flow matching to avoid the costly iterative sampling of diffusion-based approaches. The work targets low-light image enhancement (LLIE), the task of recovering well-exposed, perceptually faithful images from severely degraded observations, and positions one-step latent flow matching as a more efficient alternative to diffusion models that achieve strong perceptual quality through powerful image priors but require many sampling steps.

Low-light image enhancement LLIE aims to recover well-exposed and perceptually faithful images from severely degraded observations. Recent diffusion-based generative methods achieve strong perceptual quality by leveraging powerful image priors, but their iterative sampling incurs substantial compu