Abstract

The primary objective of the research reported in this thesis is to develop new techniques for dense point sampled surface models that directly work on the point samples. Rather than using a deterministic approach, our techniques use a statistical approach of associating properties with points based on the distribution of the point samples in a local neighborhood. Based on this research, our main thesis can be stated as follows: We can classify points using the statistical techniques of principal component analysis (PCA) into different categories such as flat, corner, crease and border. This classification can then be used to devise efficient techniques for processing such dense point sampled models. Specifically we have developed and tested new techniques for efficient rendering and reverse engineering the boundary representation of such dense point sampled models. Rendering efficiency is considerably improved by using stochastic sampling that is controlled using various model features and view dependent image space properties. (Abstract shortened by UMI.)