Refusion: Enabling Large-Size Realistic Image Restoration with Latent-Space Diffusion Models
Authors:
Top Cited Papers
- 1 June 2023
- conference paper
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 1680-1691
Abstract
This work aims to improve the applicability of diffusion models in realistic image restoration. Specifically, we enhance the diffusion model in several aspects such as network architecture, noise level, denoising steps, training image size, and optimizer/scheduler. We show that tuning these hyperparameters allows us to achieve better performance on both distortion and perceptual scores. We also propose a U-Net based latent diffusion model which performs diffusion in a low-resolution latent space while preserving high-resolution information from the original input for the decoding process. Compared to the previous latent-diffusion model which trains a VAE-GAN to compress the image, our proposed U-Net compression strategy is significantly more stable and can recover highly accurate images without relying on adversarial optimization. Importantly, these modifications allow us to apply diffusion models to various image restoration tasks, including real-world shadow removal, HR non-homogeneous dehazing, stereo super-resolution, and bokeh effect transformation. By simply replacing the datasets and slightly changing the noise network, our model, named Refusion, is able to deal with large-size images (e.g., 6000 × 4000 × 3 in HR dehazing) and produces good results on all the above restoration problems. Our Refusion achieves the best perceptual performance in the NTIRE 2023 Image Shadow Removal Challenge and wins 2 nd place overall.Keywords
All Related Versions
This publication has 43 references indexed in Scilit:
- A Style-Based Generator Architecture for Generative Adversarial Networks2019
- Scale-Recurrent Network for Deep Image Deblurring2018
- The Unreasonable Effectiveness of Deep Features as a Perceptual Metric2018
- Improving Shadow Suppression for Illumination Robust Face RecognitionIEEE Transactions on Pattern Analysis and Machine Intelligence, 2018
- Learning Deep CNN Denoiser Prior for Image Restoration2017
- Deep Residual Learning for Image Recognition2016
- Image Super-Resolution Using Deep Convolutional NetworksIEEE Transactions on Pattern Analysis and Machine Intelligence, 2015
- Improved Shadow Removal for Robust Person Tracking in Surveillance Scenarios2010
- ImageNet: A large-scale hierarchical image database2009
- Physical models for moving shadow and object detection in videoIEEE Transactions on Pattern Analysis and Machine Intelligence, 2004