IRIS Research Staff

David Page
Ph.D. Student

Office: 410 Science & Engineering
The University of Tennessee
Knoxville, TN 37996-2100
Telephone: (865) 974-9213
Fax: (865) 974-5459
E-mail: davidpage@ieee.org
Personal Web Page: /~page/


Thesis Work: Part Segmentation of Triangle Meshes

Growing consensus among cognitive psychologists suggests that the segmentation of shapes into their constituent parts is a key aid to the human visual system. Researchers such as Marr, Hoffman, and Biederman along with psychologists of the Gestalt movement argue that we see the world around us in terms of parts and that the early stages of human perception function primarily to identify features that indicate the structure of these parts. We term this visual process part-based segmentation, or more simply part segmentation. Naturally, many researchers in the computer vision community also argue that parts may be essential to computer vision tasks as well. These arguments have led us to explore part segmentation as the starting point for our research. In particular, we are interested in a specific class of part segmentation algorithms known as boundary-based methods with our main emphasis on Hoffman's minima rule.

To implement the minima rule on triangle meshes, we propose a four step process.

This output of our proposed algorithm should benefit other tasks such as scene description, object recognition, and real-time visualization.


Shape information and complexity - An algorithm that quantifies the complexity of a triangle mesh based on the entropy of curvature, which we define as shape information.


Simultaneous mesh simplification and noise smoothing - An algorithm that simultaneously simplifies the triangle count of a mesh while using normal vector voting to smooth measurement error.


Normal vector voting - An algorithm to identify surface discontinuities such as creases and corners and to estimate the curvature tensor for triangle meshes that approximate piecewise smooth surfaces.


Last updated:
Webmaster