Full Papers 5:
Rendering II
Thursday, September 1st, 2005. 08:30 - 10:30
VENUE: Burke Theatre.
SESSION CHAIR: Kari Pulli
BRDF and Geometry Capture from Extended Inhomogeneous Samples Using Flash Photography
James Paterson, David Claus, Andrew Fitzgibbon,
Oxford University Engineering Department
We present a technique which allows capture of 3D surface geometry
and a useful class of BRDFs using extremely simple equipment. A
standard digital camera with an attached flash serves as a portable
capture device, which may be used to sample geometry to very high
resolution, as well as supplying samples over a large portion of the
4D space on which the BRDF is defined. Importantly, it allows
capture of extended samples which may have spatially varying
(inhomogeneous) BRDF. We demonstrate the system by capturing the
geometry of complex materials with varying albedo and BRDF. We show
in-situ capture of materials such as a brick wall and a human hand.
The limitations of the system are that samples should be roughly
planar, and that the BRDF should have some diffuse component in
order that a first approximation to the normals can be computed.
However, given the simplicity and ease of use of the system (it
takes a few minutes to carefully capture a hand), and the ability to
capture extended surfaces without any range capture device such as a
laser scanner we argue that it is a valuable addition to the range
of real-world BRDF capture systems in the literature. We extend
standard photometric stereo techniques by moving both the camera and
the light source. By incorporating automatic parallax correction we
allow the capture of surfaces which are quite far from planar.
N-Buffers for Efficient Depth Map Query
Xavier Decoret, ARTIS GRAVIR/IMAG INRIA
We introduce the N-buffer as a tool for multiresolution depth map
representation. This neighborhood buffer encodes the value
and position of local depth extrema at different scales in an image
cube, in contrast to the image pyramid. The resulting increase in
storage space is largely compensated by the following benefits:
objects of any size can be culled in constant time against an
occlusion map using four depth lookups; visibility-like queries can
be performed in vertex and fragment programs; N-buffers can be
computed very efficiently with graphics hardware. We present three
applications of this datastructure, and in particular a novel
approach for shadow volume acceleration.
Temporally Coherent Irradiance Caching for High Quality Animation Rendering
|
Miloslaw Smyk, |
MPI Informatik and Szczecin University of Technology |
|
Shin-ichi Kinuwaki, |
Computer Graphics Laboratory, University of Aizu |
|
Roman Durikovic, |
University of Saint Cyril and Metod
|
|
Karol Myszkowski, |
MPI Informatik |
In rendering of high quality animations that include global illumination,
the final gathering and irradiance caching are commonly used. However, the
computational cost they incur is high enough to discourage their wide use
in production rendering. We introduce a data structure called anchor, which
lets us permanently link cache locations to points intersected by their
final gathering rays. Consequently, we can cheaply probe and transfer the
(ir)radiance by exploiting the temporal coherence of successive animation
frames, resulting in half an order of magnitude acceleration and reduced
temporal artifacts. Additionally, our anchor structure lets us render
moderately glossy surfaces at the cost much lower than the traditional
importance sampling techniques. We also describe an efficient, perceptually
motivated and independent scheme for limiting the growth in the number of
irradiance caches. Finally, an implementation in a practical rendering
system is demonstrated.
Spectral Volume Rendering Based on the Kubelka-Munk Theory
Alfie Abdul-Rahman, Min Chen,
University of Wales Swansea
Colour realism plays an important role in computer graphics and
visualization. In this paper, we present a new approach to direct
volume rendering based on the Kubelka-Munk theory of diffuse
reflectance. We show that not only the Kubelka-Munk theory facilitates
a correct spectral volume rendering integral suitable for both solid
objects and amorphous matters in volume datasets, but
also provides volume visualization with more accurate optical effects
than the traditional volume rendering integral based on the
RGB-alpha accumulation. We discuss the design of transfer functions
for specifying absorption and scattering coefficients, and the use of
post-illumination for integrating pre-processed reflectance images
in real time. We demonstrate the optical realism achieved by
this approach with a combination of several natural and artificial
colour datasets.
|
|
