Bayesian compressive sensing for thermal imagery using Gaussian-Jeffreys prior

Recent advances have shown a great potential to explore compressive sensing (CS) theory for thermal imaging due to the capability of recovering high-resolution information from low-resolution measurements. In this paper, we present a Bayesian CS reconstruction algorithm that makes use of a new sparsity-inducing prior, referred as Gaussian-Jeffreys prior, and demonstrate performance gain of imposing this new prior on thermal imagery where the signal-to-noise ratio is low. We first derive a hierarchical representation of the Gaussian-Jeffreys prior that facilitates computational tractability, then propose an efficient evidence approximation inference algorithm. We show that the proposed estimator is able to provide stronger sparsity-inducing power comparing to the conventional choices. Extensive numerical examples are provided with performance comparisons of different CS estimators, in particular when the compressive measurements are available via thermal imaging.