cs.CVNov 16, 2026

An Efficient Training Paradigm for Image Restoration via Improved Fréchet Loss

Authors: Shanyuan Jie, Wu Shi, Xinqi Lin, Kainan Yan, Zian Wang, Yingxue Pang, Yan Wang, Shijie Zhao, 

Organizations: Shenzhen Institutes of Advanced Technology, CAS · University of Chinese Academy of Sciences · ByteDance Inc. · The Chinese University of Hong Kong · 

Abstract

Diffusion-based image restoration achieves strong perceptual quality but typically relies on iterative sampling, leading to slow inference. Existing single-step methods often use diffusion distillation or adversarial training, complicating optimization. To simplify training, we propose FDIR, a framework for training single-pass generative restoration models through regression-style supervised optimization using directly computable objectives. To realize this paradigm, FDIR employs the Fréchet distance to match the feature distributions of restored and real images in frozen representation spaces, providing explicit distribution-level supervision. To better accommodate image restoration, we incorporate local patch features to capture fine-grained restoration discrepancies. Moreover, we introduce instance-adaptive statistics by combining the feature moments of each SR-HQ pair with synchronized historical statistics, allowing distribution matching to incorporate input-specific correspondence while retaining distribution-level context. FDIR optimizes only the LoRA parameters of the restoration backbone. Experiments demonstrate strong perceptual restoration quality and efficient training, suggesting a simpler paradigm for generative image restoration.

Figures & tables

Appendix figures & tables4 assets

Supplementary material from the paper’s appendix.

Appendix

Explore similar work

CardsList
  1. FDIR: Fidelity-Anchored Decompressed Image Restoration for Human and Machine Perception

    Jul 31, 2026Kuan-Yen Chen, Chung-Yao Ma, Fang-Yi Su +4Flow MatchingImage Restoration

  2. Noise-Free One-Step LoRA for Task-Driven Image Restoration with Diffusion Priors

    Jul 28, 2026Jaeha Kim, Kyoung Mu LeeImage RestorationTask-Driven Image Restoration

  3. TTTIR: Unlocking Instance-Specific State Evolution via Test-Time Training for Image Restoration

    Sep 15, 2026Kaihang Zheng, Jun Li, Hang Guo +5Image RestorationTest-Time Training