WaFIR: Wavelet-Fourier Mixture-of-Experts with Subband-Adaptive Routing for Blind All-in-One Image Restoration

Key points

• Blind all-in-one image restoration must adapt a shared model to heterogeneous and sometimes mixed degradations without a task label at test time.1
• Unlike image-level frequency conditioning, WaFIR makes routing decisions at the subband level, enabling frequency-specific processing while retaining the task context learned by the backbone.1
• Under a unified training and evaluation protocol for denoising, deraining, and dehazing, WaFIR improves average PSNR over the three tasks by 0.48 dB over the PromptIR baseline and by 0.38 dB over the recent Up-Restorer.1
• These results support subband frequency structure as an interpretable complement to prompt-based conditioning for blind image restoration.1

Sources (1)

• [1]WaFIR: Wavelet-Fourier Mixture-of-Experts with Subband-Adaptive Routing for Blind All-in-One Image Restoration
  arXiv (AI, ML, NLP, CV, robotics, multi-agent) · Oct 21, 02:50 AM

  We introduce WaFIR, a decoder module that decomposes features into Haar subbands, summarizes each subband with a compact Fourier descriptor, and routes it to prompt-conditioned spectral modulation experts.

  Blind all-in-one image restoration must adapt a shared model to heterogeneous and sometimes mixed degradations without a task label at test time.

Extractive summary: sentences quoted from the sources.