A universal method to evaluate and optimize any imaging system using only noisy measurements of unknown objects, enabling efficient design optimization and evaluation of real systems where traditional approaches cannot be applied.

Self-diffusion solves inverse problems without the need of pretrained generative models via a self-contained iterative process that alternates between noising and denoising steps to progressively refine its estimate of the solution.

Our work introduces a solver-agnostic null-space regularizer with gains across inverse problems.

Training PD-DL reconstruction from routine clinical images without raw data to improve fast MRI access

Whitened Score diffusion models enable stable training with arbitrary Gaussian noise by avoiding covariance inversion, improving performance on inverse problems with structured noise.