Unified descriptive experiment design regularization and component dictionary-based image restoration approach for enhanced radar/SAR sensing

The challenge of this study is to develop a new approach for multi-stage feature enhanced recovery of remote sensing (RS) imagery. The approach is based on modeling the spatial spectrum pattern (SSP) reflectivity map as a superposition of different image structures, i.e., edges, smooth and homogeneous texture zones. The latter usually manifest sparsity properties in some specific component dictionaries. The innovative proposition relates to incorporating into the initial descriptive experiment design regularization (DEDR) framework two additional regularization modalities: (i) the compressive sensing (CS) inspired convergence guaranteed regularizing projections onto convex solution sets (POCS) and (ii) the adaptive sparsity preserving despeckling level that performs the dictionary-based restoration (DBR) of the image features represented in the employed Haar wavelet dictionary basis. Algorithmically, the DBR processing is implemented as the shrinkage-type iterative CS technique adaptively incorporated into the overall multi-stage iterative DEDR-DBR method.