Efficient evaluation of functionally represented volumetric objects.

Abstract

There are several approaches to representing shapes in computer graphics. One of the ways to describe objects and operations is Function Representation (FRep). In FRep, a geometric object is de ned by a single continuous real-valued function of point coordinates. Generally geometric modelling is conducted in order to achieve visual outcome. In FRep the transformation of a function into a visual representation relies on extensive sampling of the function. The computational cost of the sampling can cause adverse e ects during applications runtime. In this thesis the problem of e cient evaluation of the de ning function is discussed. An observation is made on wide range of operations and primitives within FRep and their suitability for parallelization. Furthermore, a new novel method is proposed to distribute FReps computational workloads on parallel hardware devices such as graphics programming units and multi-core processors.