ABSTRACT
This chapter examines the problem of varying the restoration weights to handle spatially variant distortion. It analyzes the penalty function concept and shows that the method for selecting the regularization parameter based on gradient descent can find only suboptimal solutions. The chapter also shows that the natural extension to adaptive constraint restoration is to divide the image into regions, with a value of regularization parameter assigned to each region rather than each pixel. It examines the integration of the spatially variant restoration technique with the adaptive constraint technique. The chapter explores some implementation considerations and tests the algorithms developed on real imagery. It considers the use of spatially adaptive weights in constrained restoration methods. The chapter also explores how adaptive regularization techniques can compensate for insufficient knowledge of the degradation in the case of spatially variant distortions.
