Using Quadtrees to Smooth Images
Sanjay Ranade, Michael O. Shneier · 1980
Abstract : Three methods for smoothing images are presented. All three use variable numbers of picture points over which the smoothing functions are defined. The first method examines the histogram of a neighborhood of each point. The subset of gray values most similar to that of the point is used in calculating the smoothing functions. The other two methods use quadtree approximations to smooth the image. One of these uses the sizes of the leaves in the quadtree to determine neighborhood sizes over which to apply a smoothing function, while the other refines the gross smoothing defined by the quadtree. All the methods perform fairly well, but the quadtree methods are particularly attractive because of the information about region sizes and homogeneity provided by the quadtree structure.