3.6 - Regularization in Image Restoration and Reconstruction

Abstract

This chapter discusses the need for regularization in problems of image restoration and reconstruction. An overview of the issues that arise in these problems and the means to deal with them has been given in the chapter. The two driving forces in the need for regularization are noise amplification and lack of data. The primary idea behind regularization is the inclusion of prior knowledge to counteract these effects. Although there are many ways to view both these problems and their solutions, there is also a great amount of commonality in their essence. In applying such methods to image processing problems, all these approaches lead to optimization problems requiring considerable computation. Fortunately, powerful computational resources are becoming available on the desktops of nearly all engineers, and there is a wealth of complementary software tools to aid in their application. The goal of this chapter is to provide a unifying view of this area.