Restoration of images that have been blurred by the effects of a Gauss
ian blurring function is an ill-posed but well-studied problem. Any bl
ur that is spatially invariant can be expressed as a convolution kerne
l in an integral equation. Fast and effective algorithms then exist fo
r determining the original image by preconditioned iterative methods.
If the blurring function is spatially variant, however, then the probl
em is more difficult. In this work we develop fast algorithms for form
ing the convolution and for recovering the original image when the con
volution functions are spatially variant but have a small domain of su
pport. This assumption leads to a discrete problem involving a banded
matrix. We devise an effective preconditioner and prove that the preco
nditioned matrix differs from the identity by a matrix of small rank p
lus a matrix of small norm. Numerical examples are given, related to t
he Hubble Space Telescope (HST) Wide-Field/Planetary Camera. The algor
ithms that we develop are applicable to other ill-posed integral equat
ions as well.