Neural network computational technique for high-resolution remote sensing image reconstruction with system fusion

We address a new approach to the problem of improvement of the quality of scene images obtained with several sensing systems as required for remote sensing imagery, in which case we propose to exploit the idea of robust regularization aggregated with the neural network (NN) based computational implementation of the multi-sensor fusion tasks. Such a specific aggregated robust regularization problem is stated and solved to reach the aims of system fusion with a proper control of the NN's design parameters (synaptic weights and bias inputs viewed as corresponding system-level and model-level degrees of freedom) which influence the overall reconstruction performances