Spatially adaptive regularization for feature enhanced imaging in harsh sensing environments

In harsh remote sensing (RS) scenarios, reconstructive processing of the low resolution RS imagery is complicated due to the uncertainties attributed to uncontrolled multiplicative perturbations in the imaging system operators and unknown data and noise statistics. In this study, we address a new feature enhanced reconstructive RS image post processing approach that incorporates into the previously addressed unified descriptive experiment design regularization (DEDR) - variational analysis (VA) framework (DEDR-VA method) for RS image recovery in harsh sensing environments the spatially adaptive, i.e., image texture sensitive adjustments of the DEDR-VA degrees of freedom (corresponding regularization parameters). With such alternative spatially selective adaptive specifications of the DEDR-VA processing degrees of freedom, the incorporation of the third level regularization via performing projections onto convex solution sets (POCS) guarantees image sparsity and edges preservation. This enables the overall multilevel DEDR-VA-POCS (M-DEDR) technique to attain feature enhanced reconstructive imaging via performing robust spatially adaptive noise suppression with enhanced recovery of the image fine details. We feature the local statistics-based procedure for spatially adaptive adjustments of the DEDR regularization parameters and comment on the near real-time computational implementation of the overall unified M-DEDR technique. Last, the effectiveness of the proposed method is corroborated via reported computer simulation.