Shape processing defines a set of theoretical and algorithmic tools for creating, measuring and modifying digital representations of shapes. Such tools are of paramount importance to many disciplines of computer graphics, including modeling, animation, visualization, and image processing. Many applications of shape processing can be found in the entertainment and medical industries.In an attempt to improve upon many previous shape processing techniques, the present thesis explores the theoretical and algorithmic aspects of a difference measure, which involves fitting a ball (disk in 2D and sphere in 3D) so that it has at least one tangential contact with each shape and the ball interior is disjoint from both shapes.We propose a set of ball-based operators and discuss their properties, implementations, and applications. We divide the group of ball-based operations into unary and binary as follows:Unary operators include:* Identifying details (sharp, salient features, constrictions)* Smoothing shapes by removing such details, replacing them by fillets and roundings* Segmentation (recognition, abstract modelization via centerline and radius variation) of tubular structuresBinary operators include:* Measuring the local discrepancy between two shapes* Computing the average of two shapes* Computing point-to-point correspondence between two shapes* Computing circular trajectories between corresponding points that meet both shapes at right angles* Using these trajectories to support smooth morphing (inbetweening)* Using a curve morph to construct surfaces that interpolate between contours on consecutive slicesThe technical contributions of this thesis focus on the implementation of these tangent-ball operators and their usefulness in applications of shape processing.We show specific applications in the areas of animation and computer-aided medical diagnosis. These algorithms are simple to implement, mathematically elegant, and fast to execute.
PDF
download
878987797
028-85220240
