Efficient restoration of diffraction-induced space-variant blur in stacked microlens array scanning imaging system

Stacked microlens array scanning imaging systems inevitably suffer from space-variant point spread functions caused by diffraction effects, resulting in severe image blur that is difficult for traditional deconvolution methods to effectively restore. To address this, we employ a sparse matrix to efficiently represent the image degradation process and embed it into an alternating direction method of multipliers optimization framework to establish a space-variant image deblurring model. Building upon this, we design a deep unfolding network, termed the multiple space-variant kernel network (MSVKNet), and compare its restoration performance against a method based on total variation (TV) priors and conjugate gradient (CG) iteration (TV-CG). Simulation results under various system parameters indicate that MSVKNet achieves performance comparable to or better than the TV-CG method in most cases, with an inference speedup of nearly three orders of magnitude. Furthermore, experiments using a dual microlens array imaging system operating in the visible spectrum validate the accuracy and practical utility of the proposed method.