Mesh segmentation has become a key ingredient in many mesh applications in computer graphics. In this paper, we propose a hierarchical segmentation that decomposes a polygonal object into meaningful parts in such a way that not only components on a higher level reveal higher degree of salience than their descendant parts but also the components on each level of hierarchy have similar degree of salience. Moreover, the number of boundaries on each level of the hierarchy is determined automatically. The proposed segmentation is based on the Minimum Slice Perimeter (MSP) function [1], which represent non-local shape features and has better interpretation for the object parts. The gradient of MSP function is used to locate the segmentation regions and a new measure of part salience is proposed to evaluate the significance of the segmentation regions. For each level of hierarchy, some most perceptually significant segmentation regions are selected based on their salience measures and boundaries are then computed from the selected segmentation regions by using a capacity that considers both the curvature and MSP gradient.