3 - Enhancement by Multiscale Nonlinear Operators

Abstract

This chapter establishes connections between dyadic wavelet enhancement algorithms and traditional unsharp masking. It has been proved that two cases of linear enhancement are mathematically equivalent to traditional unsharp masking with Gaussian low-pass filtering. A methodology for accomplishing contrast enhancement with a simple multiscale nonlinear operator is developed that exploits the wide dynamic range available in medical images. By judicious selection of wavelet filters and enhancement function it has been shown that artifacts could be minimized. An additional advantage of this simple enhancement function is that its mathematical formulation included a generalization of traditional unsharp masking as a subset. Further, it is shown how an edge-preserving denoising stage could be incorporated into the contrast enhancement framework followed by an introduction to method for adaptive threshold estimation. Contrast enhancement was also applied to features of specific interest to mammography, including masses, spicules, and microcalcifications. Multiresolution representations provide an adaptive mechanism for the local emphasis of such features blended into digitized mammograms. In general, improvements in image contrast based on multiscale processing were superior to those obtained using competitive algorithms of unsharp masking and adaptive histogram equalization.