Near real time enhancement of geospatial imagery via systolic implementation of neural network-adapted convex regularization techniques

► A Hardware-Software (HW/SW) codesign method for enhanced sensing is addressed. ► The SW level aggregates the descriptive and Bayesian regularization paradigms. ► The HW level employs a parallelized systolic neural network computing architecture. ► Next, the HW/SW codesign is performed adapted to an FPGA processing platform. ► Attained near-real time sensing resolution enhancement is validated via simulations.