Degradation-Aware All-in-One Image Restoration via Latent Prior Encoding
This addresses the challenge of generalizing to unseen or mixed degradations in image restoration, which is important for improving visual quality and downstream vision tasks, though it appears incremental as it builds on existing all-in-one restoration frameworks.
The paper tackled the problem of all-in-one image restoration for spatially diverse degradations like haze and rain by proposing a method that learns latent priors for degradation-aware inference, resulting in an average PSNR improvement of 1.68 dB and three times greater efficiency compared to state-of-the-art approaches.
Real-world images often suffer from spatially diverse degradations such as haze, rain, snow, and low-light, significantly impacting visual quality and downstream vision tasks. Existing all-in-one restoration (AIR) approaches either depend on external text prompts or embed hand-crafted architectural priors (e.g., frequency heuristics); both impose discrete, brittle assumptions that weaken generalization to unseen or mixed degradations. To address this limitation, we propose to reframe AIR as learned latent prior inference, where degradation-aware representations are automatically inferred from the input without explicit task cues. Based on latent priors, we formulate AIR as a structured reasoning paradigm: (1) which features to route (adaptive feature selection), (2) where to restore (spatial localization), and (3) what to restore (degradation semantics). We design a lightweight decoding module that efficiently leverages these latent encoded cues for spatially-adaptive restoration. Extensive experiments across six common degradation tasks, five compound settings, and previously unseen degradations demonstrate that our method outperforms state-of-the-art (SOTA) approaches, achieving an average PSNR improvement of 1.68 dB while being three times more efficient.