Direct method speeds up solving inverse problems using diffusion models
Direct Conditional Transition Sampling for Diffusion Inverse Problems
Computer Vision and Pattern Recognition
Summary
Solving inverse problems means figuring out the original data from noisy or partial measurements, which is often slow and complex. The authors found a way to speed up this process by directly estimating certain statistics without having to do expensive intermediate calculations. Their method, called Direct Conditional Transition Sampling (DCTS), works faster while maintaining good accuracy. They tested it on several different problems and saw up to nearly 17 times faster results compared to existing approaches.
What this means in practice
- •For medical imaging teams: Speed up reconstructing images from noisy measurements to improve patient diagnostics throughput.
- •For computer vision engineers: Accelerate solving camera inverse problems like deblurring or super-resolution for real-time applications.
Authors
Qi Yu, Hanlin Wu, Xiaohui Sun
Abstract
Training-free diffusion inverse solvers typically choose between local measurement guidance and costly clean-space posterior updates. Independent posterior refresh can improve global correction by sampling a clean conditional and re-noising it, but its practical realization requires probability-flow ODE integration and clean-space Markov chain Monte Carlo (MCMC). We propose Direct Conditional Transition Sampling (DCTS), a direct stochastic-flow approximation to the same ideal refresh target. Rather than explicitly drawing a clean sample, DCTS estimates the measurement-conditioned clean mean along a short inner path and transports Gaussian source noise directly to the next noisy state. A denoiser-compatible sufficient statistic and a covariance-scaled operator update enable this conditional-mean estimation. Experiments on four inverse problems demonstrate that DCTS achieves competitive reconstruction quality with up to $16.8\times$ speedups over competing methods.