Coupling Model- and Data-Driven Methods for Remote Sensing Image Restoration and Fusion: Improving physical interpretability

In the fields of image restoration and image fusion, model- and data-driven methods are the two representative frameworks. However, both approaches have their respective advantages and disadvantages. Model-driven techniques consider the imaging mechanism, which is deterministic and theoretically reasonable; however, they cannot easily model complicated nonlinear problems. Data-driven schemes have a stronger prior-knowledge learning capability for huge data, especially for nonlinear statistical features; however, the interpretability of the networks is poor, and they are overdependent on training data. In this article, we systematically investigate the coupling of model- and data-driven methods, which has rarely been considered in the remote sensing image restoration and fusion communities. We are the first to summarize the coupling approaches into the following three categories: 1) data- and model-driven cascading methods, 2) variational models with embedded learning, and 3) model-constrained network learning methods. The typical existing and potential coupling techniques for remote sensing image restoration and fusion are introduced with application examples. This article also gives some new insights into potential future directions, in terms of both methods and applications.