Spatio-temporal resolution enhancement for geostationary microwave data

In this paper, we provide a formulation for enhancing the spatio-temporal resolution of a remote sensing sequence of images. Such an image sequence could be captured by a sensor that convolves a physical scene with a spatio-temporal point spread function whose two-dimensional spatial component is the microwave instrument's point spread function and whose one-dimensional temporal component is the rectangular kernel with sensor exposure time as its support. We perform resolution enhancement in the space-time domain, as opposed to solving the deconvolution problem for each observation. Simultaneous space-time optimization achieves a more efficient and more accurate reconstruction. The proposed deconvolution method employs total variation regularization and solves the formulation via the Split-Bregman optimization algorithm. In our experiments, we use a simulated microwave image sequence of a hurricane and demonstrate that the proposed methodology improves the accuracy when compared to the observed sequence.