Come-Closer-Diffuse-Faster: Accelerating Conditional Diffusion Models for Inverse Problems through Stochastic Contraction
Authors:
Top Cited Papers
- 1 June 2022
- conference paper
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 12403-12412
Abstract
Diffusion models have recently attained significant interest within the community owing to their strong performance as generative models. Furthermore, its application to inverse problems have demonstrated state-of-the-art performance. Unfortunately, diffusion models have a critical downside - they are inherently slow to sample from, needing few thousand steps of iteration to generate images from pure Gaussian noise. In this work, we show that starting from Gaussian noise is unnecessary. Instead, starting from a single forward diffusion with better initialization significantly reduces the number of sampling steps in the reverse conditional diffusion. This phenomenon is formally explained by the contraction theory of the stochastic difference equations like our conditional diffusion strategy - the alternating applications of reverse diffusion followed by a non-expansive data consistency step. The new sampling strategy, dubbed Come-Closer-Diffuse-Faster (CCDF), also reveals a new insight on how the existing feed-forward neural network approaches for inverse problems can be synergistically combined with the diffusion models. Experimental results with super-resolution, image inpainting, and compressed sensing MRI demonstrate that our method can achieve state-of-the-art reconstruction performance at significantly reduced sampling steps.Keywords
All Related Versions
Funding Information
- National Research Foundation(NRF) of Korea (NRF-2021M3I1A1097938)
This publication has 10 references indexed in Scilit:
- ILVR: Conditioning Method for Denoising Diffusion Probabilistic Models2021
- Photo-Realistic Single Image Super-Resolution Using a Generative Adversarial Network2017
- Projections onto Convex Sets Super-Resolution Reconstruction Based on Point Spread Function Estimation of Low-Resolution Remote Sensing ImagesSensors, 2017
- Projection onto convex sets super-resolution image reconstruction based on wavelet bi-cubic interpolation2011
- Convex Analysis and Monotone Operator Theory in Hilbert Spaces2011
- ImageNet: A large-scale hierarchical image database2009
- A Contraction Theory Approach to Stochastic Incremental StabilityIEEE Transactions on Automatic Control, 2009
- Analysis of discrete and hybrid stochastic systems by nonlinear contraction theory2008
- Undersampled radial MRI with multiple coils. Iterative image reconstruction using a total variation constraintMagnetic Resonance in Medicine, 2007
- POCSENSE: POCS‐based reconstruction for sensitivity encoded magnetic resonance imagingMagnetic Resonance in Medicine, 2004