This story is from the category Libraries and Components
Date posted: 19/03/2007
Research coming out of the University of Saarland in Germany promises to revolutionise the speed of raytracing algorithms, making utterly realistic light and shadow available to even the slowest pcs.
The knock-on implications for VR are simply staggering.
Usually, for raytracing of the extent seen in pre-rendered images, it took hours per frame, or special, heavy-duty hardware to render. But, the new method, can be done on any pc, with a plug-in card of a custom chipset, or even the power of a high-end graphics card ? irrespective of the capabilities of the rest of the system.
Real-time graphics are typically rendered via a technique known as rasterisation which involves drawing all the elements of a scene using polygons. Raytracing is vastly superior, because it exactly models the passage of light from every light source to the viewer?s eye, giving every visible surface the correct blend of light and shadow.
However the computational power needed to keep track of the light has meant that it can only be done using lots and lots of time per individual frame, or a whole, dedicated cluster server.
Professor Philipp Slusallek and co-workers from the University of Saarland have developed a series of new ray-tracing algorithms that promise to make it much easier to use the technique.
Daniel Pohl, one of the researchers who has worked with Professor Slusallek, has used the algorithms to produce ray-traced versions of the Quake 3 and 4 video games.
"It gives much higher image quality in shadows and reflections," said Mr Pohl. "You can even do reflections on reflections on reflections."
Straight away, we therefore have effects, hitherto impossible to compute graphically, being performed on mid-end PCs. On top of that, the way the algorithms work, is ideally suited to multi-core processing.
The algorithms are being made available to anyone to use via the Open RT project.
See the full Story via external site: news.bbc.co.uk
Most recent stories in this category (Libraries and Components):
17/02/2015: New algorithms Geolocate a video from its images and sounds