Hierarchical image segmentation using a watershed scale-space tree

The watershed transformation is a useful morphological segmentation tool which has been used in a variety of grey-scale image processing applications. However, a major problem with the watershed transformation is that it produces a severe over-segmentation due to the great number of minima embedded in the image or its gradient, and therefore it is rarely applied directly to images. In this paper we discuss modifications to the basic watershed transformation that enable watershed scale trees to be produced and we illustrate the approach with some example segmentations taken from a medical image processing application. The mapping between image and scale tree allows the user to overcome the over-segmentation by either setting a global threshold or interactively editing the tree description of the image.