Regularization of deconvolution using steerable pyramids

In this paper, a method is proposed to regularize deconvolution based on steerable pyramids. Deconvolution is the process of recovering an ideal image from an observation, usually degraded by blurring and a certain noise process (additive, multiplicative, ...). Since deconvolution is an ill-posed problem, noise will be amplified when no regularization is applied. This regularization step formulates our prior knowledge about the image, like general smoothness, ... We propose a regularization approach where deconvolution is combined with a noise reduction scheme, based on steerable pyramids.