Block Coordinate Plug-and-Play Methods for Blind Inverse Problems

Authors: Weijie Gan, shirin shoushtari, Yuyang Hu, Jiaming Liu, Hongyu An, Ulugbek Kamilov

NeurIPS 2023 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We numerically validate our method on two blind inverse problems: automatic coil sensitivity estimation in magnetic resonance imaging (MRI) and blind image deblurring. Our results show that BC-Pn P provides an efficient and principled framework for using denoisers as Pn P priors for jointly estimating measurement operators and images.
Researcher Affiliation Academia Weijie Gan Washington University in St. Louis; Shirin Shoushtari Washington University in St. Louis; Yuyang Hu Washington University in St. Louis; Jiaming Liu Washington University in St. Louis; Hongyu An Washington University in St. Louis; Ulugbek S. Kamilov Washington University in St. Louis
Pseudocode Yes Algorithm 1 Block Coordinate Plug-and-Play Method (BC-Pn P)
Open Source Code Yes Our code which we share publicly shows the potential of learning deep denoisers over measurement operators and using them for jointly estimating the uknown image and the uknown measurement operator.
Open Datasets Yes We used T2-weighted MR brain acquisitions of 165 subjects obtained from the validation set of the fast MRI dataset [75] as the the fully sampled measurement for simulating measurements... We randomly selected 10 testing ground truth image from CBSD68 [80] dataset.
Dataset Splits Yes These 165 subjects were split into 145, 10, and 10 for training, validation, and testing, respectively.
Hardware Specification No The paper describes the experimental setup and training procedures but does not specify any particular hardware components such as GPU or CPU models.
Software Dependencies No The paper mentions deep learning architectures like DRUNet [12] and Dn CNN but does not provide specific version numbers for software dependencies or libraries.
Experiment Setup No The paper describes the initialization of BC-Pn P and ablated variants, and the architectures used for denoisers (DRUNet, Dn CNN). However, it does not explicitly state key hyperparameters like learning rate, batch size, or number of epochs in the main text.