A Method for Hierarchical Description of Images Using Multiscale Analysis

Abstract

The aim of this paper is to present a novel method of image segmentation based on multiresolution analysis. The main idea consists of the extraction of image parts which are coherent and are enclosed by a coherent region classified to another type. Segmentation is processed in two steps. The first, consists of pixel classification into predefined types and the second is detection of the hierarchical structure of an image. The method is presented here on examples being biomedical images but it may be also applied to analyze other types of images.