Advances in Image Restoration and Enhancement

The field of image restoration and enhancement is rapidly evolving, with a focus on developing innovative methods that can effectively address various challenges such as low-light conditions, degradation, and noise. Recent research has explored the use of diffusion models, deep reinforcement learning, and latent space representations to improve image quality. Notably, the integration of diffusion training paradigms into general image restoration frameworks has shown promising results, enabling simultaneous image restoration and generative representation modeling. Furthermore, personalized low-light image enhancement methods and lightweight blind super-resolution models have demonstrated superior performance and adaptability.

Noteworthy papers include: Elucidating and Endowing the Diffusion Training Paradigm for General Image Restoration, which proposes a new framework for adapting diffusion training paradigms to general image restoration tasks. ReF-LLE: Personalized Low-Light Enhancement via Reference-Guided Deep Reinforcement Learning, which introduces a novel personalized low-light image enhancement method that operates in the Fourier frequency domain and incorporates deep reinforcement learning. LightBSR: Towards Lightweight Blind Super-Resolution via Discriminative Implicit Degradation Representation Learning, which proposes a lightweight blind super-resolution model that focuses on the discriminability optimization of implicit degradation representation.

Sources

Elucidating and Endowing the Diffusion Training Paradigm for General Image Restoration

ReF-LLE: Personalized Low-Light Enhancement via Reference-Guided Deep Reinforcement Learning

LightBSR: Towards Lightweight Blind Super-Resolution via Discriminative Implicit Degradation Representation Learning

UniFuse: A Unified All-in-One Framework for Multi-Modal Medical Image Fusion Under Diverse Degradations and Misalignments

Degradation-Modeled Multipath Diffusion for Tunable Metalens Photography

Efficient Depth- and Spatially-Varying Image Simulation for Defocus Deblur

Latent Posterior-Mean Rectified Flow for Higher-Fidelity Perceptual Face Restoration

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

DiffusionLight-Turbo: Accelerated Light Probes for Free via Single-Pass Chrome Ball Inpainting

DocShaDiffusion: Diffusion Model in Latent Space for Document Image Shadow Removal

Posterior Transition Modeling for Unsupervised Diffusion-Based Speech Enhancement

PosDiffAE: Position-aware Diffusion Auto-encoder For High-Resolution Brain Tissue Classification Incorporating Artifact Restoration

IGDNet: Zero-Shot Robust Underexposed Image Enhancement via Illumination-Guided and Denoising

Built with on top of