A FLEXIBLE IMPLEMENTATION OF MAXIMUM-HOMOGENEITY FILTERING FOR 2-D AND 3-D IMAGES

Nonlinear edge-preserving filtering techniques have long been popular in image processing. One of these techniques, the maximum-homogeneity filter, has shown promise for reducing random noise, sharpening blurred regional borders, and removing unwanted thin artifacts. Unfortunately, previous authors have only defined the maximum-homogeneity filter for fixed circumstances. We define, for both three-dimensional and two-dimensional images, a more flexible implementation of the maximum-homogeneity filter. The implementation, based on a mathematically defined "spotlight" operator, permits varying levels of region sharpening and artifact removal. Test results demonstrate that the technique gives performance comparable to other edge-preserving filters.