Reconstruction of multimode remote sensing imagery based on multilevel experiment design regularization

In this paper, we propose a new multilevel regularization technique for solving nonlinear inverse problems of reconstruction of remote sensing (RS) imagery acquired with low resolution radar sensor systems of different modalities operated in the scenarios with perturbed (uncertain) system operators. We infer the proposed generalized multilevel regularization from the unified experiment design regularization (EDR) framework based on the conditioned minimum risk reconstruction strategy. The first two regularization levels perform the range-azimuth factorization and low resolution image despeckling adapted to the particular uncertain scenario; the third level balances the image gradient sparsity over the homogeneous scene regions with edge preservation, and at the fourth level, the projection onto positive convex solution set is incorporated to speed-up the resulting parallelized contractive mapping iterative RS image reconstruction procedure. The addressed multilevel EDR (M-EDR) technique endows the previously addressed EDR approaches to cope with the perturbed RS imaging models exploiting in the same time the image model sparsity combined with structured parallelized implementation. The simulations demonstrate the effectiveness of the developed technique in comparison with the recently proposed most competing nonparametric radar imaging approaches that do not employ the M-EDR.