Research Projects
2025 Fall (Current)
- Variational Inference with Diffusion Priors: This project aims to investigate approaches to combine variational inference models (e.g., VAE) with diffusion models that serve as an informative prior. Applications will include image restoration tasks such deblurring, inpainting, and super-resolution.
Status: available
Contact: Evan Bell (ebell34@jh.edu)
- Diffusion Models for Synthesizing RFO-contained Intraoperative Radiographs: Retained foreign objects (RFOs)—including sponges, needles, sutures, and other instruments—can cause severe harm to patients. This project aims to synthesize intraoperative radiographs containing RFOs to enhance their detection.
Status: available
- Enhancing Brian MRI using Diffusion Model: MRI has become an indispensable tool for measuring the blood flow in the brain for diagnosis (e.g. stroke). However, current scan suffers from low SNR and long acquisition time. The goal of this project is to develop novel algorithms to enhance the MRI scan quality.
Status: reserved (Haoyue Guan)
- Flow Matching for Super-Resolution in Optics: Flow matching is a novel generative model that outperforms diffusion models. This project aims to explore the use of flow matching for super-resolving optical images.
Status: reserved (Yetao He)
2025 Summer
Neural Inverse Scattering with Score-based Regularization (Yuan Gao, Yu Sun)
Blind Deblurring using Measurement-Conditioned Diffusion Priors (Yuanyun Hu, Yu Sun)
2024 Fall
Neural Inverse Scattering with Score-based Regularization (Yuan Gao, Yu Sun)
Blind Deblurring with Application to Optometry (Guannan He, Yuanyun Hu, Yu Sun)
Exploration of Semantic Uncertainty in Diffusion Models (Xinyuan Shao, Yu Sun)
Design of Efficient Plug-and-Play Algorithms (Bingyan Liang, XInmin Shen, Yu Sun)
© Copyright 2025 Hopkins Computational Imaging
Electrical and Computer Engineering
Department
3400 North Charles Street
Baltimore, MD 21218