3.5 - Basic Methods for Image Restoration and Identification

Abstract

The image restoration methods that are described in this chapter fall under the class of linear spatially invariant restoration filters. It has been assumed that the blurring function acts as a convolution kernel or a point-spread function that does not vary spatially. It is also assumed that the statistical properties of the image and noise do not change spatially. Under these conditions the restoration process can be carried out by means of a linear filter, of which the point-spread function is spatially invariant—that is, it is constant throughout the image. In addition, important models for linear blurs—namely, motion blur, out-of-focus blur, and blur because of atmospheric turbulence—are described. Three classes of restoration algorithms are also introduced and described in the chapter: the inverse filter, the Wiener and constrained least-squares filter, and the iterative restoration filters.