COMP-766 Shape Analysis in Computer Vision Project Report MATCHING SHAPES BY PROBABILISTIC GRAPH MATCHING

The ability to recognize shape is crucial in many vision applications. If a given shape could be accurately modelled and matched to a base shape with some degree of certainty, then the a host of applications can be developed from computer vision to medical imaging. This project analyses some newly developed ideas which seek to extend well known notions of shape modelling by skeleton and probabilistic inexact graph matching. Further, issues will be discussed on how best to represent the shape as a graph by skeletal means and how a probabilistic Bayesian network can match these graphs.