Limited-memory scaled gradient projection methods for real-time image deconvolution in microscopy

摘要

Gradient projection methods have given rise to effective tools for image deconvolution inudseveral relevant areas, such as microscopy, medical imaging and astronomy. Due to theudlarge scale of the optimization problems arising in nowadays imaging applications andudto the growing request of real-time reconstructions, an interesting challenge to be facedudconsists in designing new acceleration techniques for the gradient schemes, able toudpreserve their simplicity and low computational cost of each iteration. In this work we proposeudan acceleration strategy for a state-of-the-art scaled gradient projection method forudimage deconvolution in microscopy. The acceleration idea is derived by adapting a steplengthudselection rule, recently introduced for limited-memory steepest descent methodsudin unconstrained optimization, to the special constrained optimization framework arisingudin image reconstruction. We describe how important issues related to the generalization ofudthe step-length rule to the imaging optimization problem have been faced and we evaluateudthe improvements due to the acceleration strategy by numerical experiments onudlarge-scale image deconvolution problems.