Interleaved Sampling
Alexander
Keller, Wolfgang Heidrich
Introduction
On this web page we illustrate the quality
benefits of interleaved sampling more clearly than this is possible in
a printed document. Especially in the volume rendering part we also provide
some additional material that is not in the printed Rendering
Workshop paper.
Overview:
Abstract
The sampling of functions is one of the most
fundamental tasks in computer graphics, and occurs in a variety of different
forms. The known sampling methods can roughly be grouped in two categories.
Sampling on regular grids is simple and efficient, and the algorithms are
often easy to built into graphics hardware. On the down side, regular sampling
is prone to aliasing artifacts that are expensive to overcome. Monte Carlo
methods, on the other hand, mask the aliasing artifacts by noise.
However, due to the lack of coherence known from regular grids, and are
more expensive and not well suited for hardware implementations.
In this paper, we introduce a novel sampling
scheme where samples from several regular grids are a combined into a single
irregular sampling pattern. The relative positions of the regular grids
are themselves determined by Monte Carlo methods. This generalization obtained
by interleaving yields significantly improved quality compared to traditional
approaches while at the same time preserving much of the advantageous coherency
of regular sampling.
We demonstrate the quality of the new sampling
scheme with a number of applications ranging from supersampling over effects
like shadow maps, depth-of-field and motion blur to volume rendering. Due
to the coherence in the interleaved samples, the method is optimally suited
for implementations in graphics hardware.