Federal Grant Award Summary

The National Science Foundation's Division of Computing and Communication Foundations awarded The Johns Hopkins University a $441,528 Project Grant (effective April 15, 2026 through March 31, 2031) under the Computer and Information Science and Engineering (CISE) program (CFDA 47.070) to develop novel computational imaging methods. The project will deliver a provable, flexible, and scalable framework that integrates physical forward models with generative diffusion models to enable probabilistic imaging approaches. Rather than producing single-image reconstructions, the framework will recover the full distribution of plausible image solutions consistent with measured data, thereby providing principled uncertainty characterization—critical for scientific and clinical decision-making in applications such as three-dimensional microscopy.

The research will pursue three integrated technical directions: establishing a rigorous posterior sampling framework for imaging inverse problems with theoretical guarantees; designing flexible algorithms compatible with nonlinear and partially unknown forward models; and creating scalable methods for high-dimensional imaging applications. These deliverables address a significant gap in current computational imaging practice by shifting from deterministic to probabilistic reconstruction approaches that can characterize ambiguity and indicate trustworthiness of recovered images. The award supports foundational and applied research aligned with CISE's mission to advance discovery and innovation across computing and information science and engineering domains.