Use of blur-space for deblurring and edge-preserving noise smoothing

Abstract

The Gaussian blur-space for an unblurred nD-image I is the set of the images obtained by blurring I with multivariate nD-Gaussians. Using the variance, instead of the standard deviation, of a Gaussian as blur parameter makes it simpler to extrapolate a deblurred image from a blurred image. Unsharp masking is shown to be a special case of the use of blur-space. Algorithms using blur-space for deblurring and edge-preserving noise smoothing, without explicit edge detection, are described and implemented.