Image segmentation by a multiresolution approach

Abstract Multiresolution approaches to computer vision are able to rapidly detect and extract global structures from an image. In this paper we present (a) a pyramid-based algorithm that can detect the bimodality of the population of pixels in a grey level digital image and (b) a pyramid-based algorithm that maps the values of a bimodal population into two constant values which are approximately the means of the two component subpopulations. A population is considered bimodal if it can be divided into two component subpopulations whose variances are small relative to the population variance. An improvement to the above algorithm, which uses an iterative scheme, is also given, as well as some examples of segmented images. Both algorithms require processing times on the order of the logarithm of the population size.