Restoration of bilinearly distorted images 3: Constrained least-squares method

Abstract

A constrained least-squares solution based on a second difference smoothing measure is obtained for images which are distorted by a bilinear (quadratic and with nonzero memory) system in the presence of additive signal-independent noise. Results are applied to a partially coherent diffraction-limited imaging system. It is found that the optimum weight given to the smoothness factor is larger for a coherent (nonlinear) system than for an incoherent (linear) system. However, the restoration quality improves as the imaging system approaches incoherence.