Hongjiang Wei

By explicitly modeling the unknown forward operator with a monotonic constraint, this work achieves state-of-the-art results in blind inverse problems like low-light enhancement and HDR reconstruction.

Achieve state-of-the-art low-field MRI enhancement by explicitly modeling and correcting the intensity distribution shift between low-field and high-field domains using a differentiable Sinkhorn optimal transport module within a diffusion framework.

By restoring the classical dual variable to Plug-and-Play Diffusion, this work guarantees asymptotic convergence to the exact data manifold in medical image reconstruction, resolving the bias-hallucination trade-off that plagues existing methods.