Shape Analysis - Duke CPS 296.2 - Spring 2004
Pankaj K. Agarwal
Phone: (919) 660-6540
The course is intended to provide a systematic introduction to the modeling and algorithmic techniques behind the geometric and statistical analysis of 3D shapes that arise in many applications including molecular biology, computer graphics, and computer aided design. The primary emphasis will be on recent algorithms developed for representing, analyzing, comparing, classifying, and indexing 3D shapes. The topics covered will include:
I. Shape representation: basic representation methods, shape simplification, hierarchical methods, deformable shapes.
II. Shape descriptors: histograms, harmonic maps, distance distribution, medial axis, topology based methods.
III. Statistical shape analysis: Shape space, coordinate systems, procrustes distances and their generalizations, deformations.
IV. Shape matching and registration: Combinatorial methods, geometric hashing, ICP and its variants, graph matching, entropy based methods.
V. Shape classification and clustering: Geometric clustering, graph based methods, spectral methods, decision trees, support vector machines.
VI. Shape indexing: Indexing multidimensional data, proximity search, search engines.
The course is open to graduate and undergraduate students from all fields.
Assignments: 20% weight
Two assignments will be given during the semester, which each student has to complete individually.
Research Project: 40% weight
The intention is to produce a work of publishable or near-publishable quality. It will consist of two parts:
(i) proposal and survey
(ii) research work (a final paper and presentation)

Lectures: 40% weight
Each student will be asked to give one or two lectures, which will require reading papers on a given topic and presenting the technqiues and results.

Web site by Celeste Hodges