Extension of the deep equilibrium framework to solve linear inverse problems in computational imaging.
Code
Neumann networks are a neural network meta-architecture that can be trained to solve linear inverse problems arising in computational imaging.
Link to Project Page Page
Code
An extension of low-rank matrix completion to the case where the data lies on a low-dimensional algebraic variety. This includes data belonging to a union of low-dimensional subspaces, and other low-dimensional algebraic curves and surfaces.
Variety Matrix Completion (VMC) Code
Low Algebraic Dimension Matrix Completion (LADMC) Code
A fast algorithm for solving low-rank matrix completion problems with additional (block)-Toeplitz or (block)-Hankel structure, with application to undersampled MRI reconstruction.
Code
An extension of the finite-rate-of-innovation (FRI) framework to a class of multi-dimensional piecewise smooth image models.
Code
An efficient extension of the total variation regularization penalty to higher order derivatives for use in inverse problems in imaging.
Link to CBIG Project Page
Code
An algorithm for MRI reconstruction that utilizes a patch-based similarity prior.
Link to CBIG Project Page