The field of image restoration is rapidly advancing with the development of diffusion models. These models have shown great promise in solving inverse problems, such as image inpainting and super-resolution, by progressively transforming pure noise into structured data through a denoising process. Recent research has focused on improving the efficiency and accuracy of diffusion models, including the use of piecewise guidance schemes, latent diffusion-enhanced vector-quantized codebook priors, and adaptive path tracing. Noteworthy papers in this area include Diffusion Models for Solving Inverse Problems via Posterior Sampling with Piecewise Guidance, which introduces a novel diffusion-based framework for solving inverse problems, and UniLDiff: Unlocking the Power of Diffusion Priors for All-in-One Image Restoration, which proposes a unified image restoration framework based on latent diffusion models. Additionally, papers such as MoFRR: Mixture of Diffusion Models for Face Retouching Restoration and All-in-One Medical Image Restoration with Latent Diffusion-Enhanced Vector-Quantized Codebook Prior demonstrate the effectiveness of diffusion models in specific applications, including face retouching restoration and medical image restoration.
Advances in Image Restoration with Diffusion Models
Sources
Diffusion Models for Solving Inverse Problems via Posterior Sampling with Piecewise Guidance
Plug and Play Splitting Techniques for Poisson Image Restoration
DEFNet: Multitasks-based Deep Evidential Fusion Network for Blind Image Quality Assessment
An inverse random diffraction grating problem for the Helmholtz equation
MoFRR: Mixture of Diffusion Models for Face Retouching Restoration
All-in-One Medical Image Restoration with Latent Diffusion-Enhanced Vector-Quantized Codebook Prior
Fine-structure Preserved Real-world Image Super-resolution via Transfer VAE Training
Conditional Diffusion Models for Global Precipitation Map Inpainting
Harnessing Diffusion-Yielded Score Priors for Image Restoration
APT: Improving Diffusion Models for High Resolution Image Generation with Adaptive Path Tracing
Exploiting Diffusion Prior for Task-driven Image Restoration
LCS: An AI-based Low-Complexity Scaler for Power-Efficient Super-Resolution of Game Content
UniLDiff: Unlocking the Power of Diffusion Priors for All-in-One Image Restoration