This repository provides official implementations of IR-SDE and Refusion, two diffusion-based image restoration methods. IR-SDE uses mean-reverting stochastic differential equations to model the restoration process, while Refusion enables large-size realistic image restoration using latent-space diffusion models. Both approaches handle tasks including deblurring, denoising, super-resolution, deraining, dehazing, inpainting, and shadow removal. The methods are trained on synthetic and real-world datasets and achieved state-of-the-art results, winning the NTIRE 2023 Image Shadow Removal Challenge.