Diploma/Master's Thesis:

"Photon Density Estimation Using Wavelet Compression"

Christian Lauterbach (cl (at) cs.unc.edu)

Universität Bremen, 2005

Advisors:
 Frieder Nake 
 Friedrich-Wilhelm Bruns


Abstract:

The physically-based simulation of light transport in synthetic scenes is an area in computer graphics
known as global illumination and its most important application is the creation of 'realistic' images.
As this is of obvious interest, many algorithms were developed for this purpose. However, rendering
images is still often a slow process and hardly interactive.

This thesis presents an interesting variant on existing two-pass global illumination
algorithms that use so-called density estimation to calculate the illumination
at points in the scene. There are two main advantages of this algorithm: first,
determining the illumination for rendering has essentially constant time complexity
and is independent of the detail of the solution. This means that the quality of the illumination,
which is calculated in a precomputation step, can be increased without slowing
the subsequent rendering.

Second, since the algorithm allows a very exact precomputation,
the complete illumination (direct and indirect) can be taken from the precomputed
solution, which reduces the rendering step to ray-casting and illumination look-up only.
This may not be as fast as algorithms that produce an illuminated mesh that can be
displayed in real-time using graphics hardware, but the rendering system implementing
the algorithms discussed in this thesis employs a ray-tracing optimization that enables
to trace several rays in parallel which speeds up the rendering process by a factor of
two. While this does not yet allow interactive walkthrough, at least navigation in the scene and
display of the global illumination solution is possible using a viewer application.


Download:
 - PDF
 - PDF (without colored links, suitable for printing)