Hierarchical Radiosity with Multiresolution Meshes
Andrew J. Willmott
CMU-CS-00-166
Thesis Committee:
Paul Heckbert, Chair
David O'Hallaron
Steven Seitz
François Sillion
(c) 2000 by Andrew Willmott
The hierarchical radiosity algorithm solves for the global
transfer of diffuse illumination in a scene. While its potential algorithmic
complexity is superior to both previous radiosity methods and distributed
ray tracing, for scenes containing detailed polygonal models, or highly tessellated
curved surfaces, its time performance and memory consumption are less than
ideal.
My thesis is that by using
hierarchies similar to those of multiresolution models, the performance of
the hierarchical radiosity algorithm can be made sub-linear in the number
of input polygons, and thus make radiosity on scenes containing detailed models
tractable. The underlying goal of my thesis work has been to make high-speed
radiosity solutions possible with such scenes.
To achieve this goal,
a new face clustering technique for automatically partitioning polygonal models
has been developed. The face clusters produced group adjacent triangles with
similar normal vectors. They are used during radiosity solution to represent
the light reflected by a complex object at multiple levels of detail. Also,
the radiosity method is reformulated in terms of vector irradiance. Together,
face clustering and the vector formulation of radiosity permit large savings.
Excessively fine levels of detail are not accessed by the algorithm during
the bulk of the solution phase, greatly reducing its memory requirements
relative to previous methods. Consequently, the costliest steps in the simulation
can be made sub-linear in scene complexity.
I have developed a radiosity
system incorporating these ideas, and shown that its performance is far superior
to existing hierarchical radiosity algorithms, in the domain of scenes containing
complex models.
Online Version
The online version is available as the original postscript, and two PDF files.
The smaller PDF file was produced by using the default Adobe Distiller settings.
The larger file was produced by forcing Distiller to use lossless compression
(zip) for all images. If you have the bandwidth, the larger file is recommended,
as many images in the smaller PDF contain objectionable JPEG artifacts.
Individual Chapters
The following files are in PDF format, with lossless compression only.
Chapter
2
Previous Work on the Radiosity Problem
Chapter
5
Improving the Face Cluster Construction Algorithm
References
Defense Slides
Slides from my oral presentation.
Source Code
The source code that implements the methods described in the dissertation
is being released at
http://www.cs.cmu.edu/~ajw/thesis-code/.
ajw+thesis@cs.cmu.edu