Model Adaptation for Inverse Problems in Imaging
This work is significant for researchers and practitioners using deep learning for inverse problems in imaging, as it offers a solution to the common challenge of model degradation due to forward model mismatches, potentially reducing the need for extensive retraining.
This paper addresses the problem of deep neural networks' sensitivity to changes in the forward model for inverse problems in imaging. The authors propose two novel procedures to adapt a pre-trained network to a new forward model without requiring additional labeled data, demonstrating empirical success across various inverse problems like deblurring and super-resolution.
Deep neural networks have been applied successfully to a wide variety of inverse problems arising in computational imaging. These networks are typically trained using a forward model that describes the measurement process to be inverted, which is often incorporated directly into the network itself. However, these approaches are sensitive to changes in the forward model: if at test time the forward model varies (even slightly) from the one the network was trained for, the reconstruction performance can degrade substantially. Given a network trained to solve an initial inverse problem with a known forward model, we propose two novel procedures that adapt the network to a change in the forward model, even without full knowledge of the change. Our approaches do not require access to more labeled data (i.e., ground truth images). We show these simple model adaptation approaches achieve empirical success in a variety of inverse problems, including deblurring, super-resolution, and undersampled image reconstruction in magnetic resonance imaging.