Posterior sampling

We propose a method to perform posterior sampling with diffusion models on blind inverse problems.

Diffusion posterior sampling enables solving arbitrary noisy (e.g. Gaussian, Poisson) inverse problems that are both linear or non-linear.