AI Paper Atlas

Research map / Image generation / Image restoration

LD-RPS: Zero-Shot Unified Image Restoration via Latent Diffusion Recurrent Posterior Sampling

Huaqiu Li, Yong Wang, Tongwen Huang, Hailang Huang, Haoqian Wang, Xiangxiang Chu

ICCV 2025 · Image generation › Image restoration · approach: Inverse imaging with learned priors · 8 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
add MLLM text guidance (ablation)LD-RPS · LOLv1 (low-light enhancement)PSNR: 16.03 -> 17.73Improved
add MLLM text guidance (ablation)LD-RPS · RESIDE HSTS (dehazing)PSNR: 19.63 -> 21.60Improved
add MLLM text guidance (ablation)LD-RPS · Kodak24 (denoising)PSNR: 28.13 -> 28.60Improved
add MLLM text guidance (ablation)LD-RPS · RESIDE HSTS (dehazing)LPIPS: 0.169 -> 0.179Hurt
add recurrent refinement (2 recurrences vs 0) (ablation)LD-RPS · LOLv1 (low-light enhancement)PSNR: 16.78 -> 17.73Improved
increase recurrences from 2 to 3 (ablation)LD-RPS · LOLv1 (low-light enhancement)PSNR: 17.73 -> 17.10Hurt
add recurrent refinement (3 recurrences vs 0) (ablation)LD-RPS · RESIDE HSTS (dehazing)PSNR: 19.35 -> 21.60Improved
add recurrent refinement (1 recurrence vs 0) (ablation)LD-RPS · Kodak24 (denoising)PSNR: 27.75 -> 28.60Improved
increase recurrences from 1 to 3 (ablation)LD-RPS · Kodak24 (denoising)PSNR: 28.60 -> 28.49No clear effect
use LD-RPS instead of TAOposterior sampling · LOLv1 (low-light enhancement)PSNR: 15.84 -> 17.45Improved
use LD-RPS instead of TAOposterior sampling · RESIDE HSTS (dehazing)PSNR: 18.38 -> 21.45Improved
use LD-RPS instead of TAOposterior sampling · RESIDE HSTS (dehazing)SSIM: 0.823 -> 0.813Hurt

Later papers that compared with this method

1 later accepted paper reports LD-RPS as a baseline in their results tables; 1 of them report a better number on most of the results they share with it.

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.

Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy