bidir2.avi

Fast bidirectional path tracer working on a simple scene. Screen resolution is 1024x512, number of rays per second is close to 20M/s. The demo is running on a core i7 with NVidia GTX260 graphics. Code by Jacco Bikker & Dietger van Antwerpen. More information on http://igad.nhtv.nl/~bikker .

Category:

Science & Technology

Tags:

• RayGrid
• 2010
• 02
• 08
• 14
• 37
• 19
• 41

[robinbuster]

amazing!

[edwardkmett]

@rickb1988nl Ah fair enough. Say Hi to Jacco for me. We used to chat quite a bit back in the flipcode era when I went by Harmless.

[rickb1988nl]

@phoenix13nl No,I mean the jpeg blocks you see in the video as it converges to a less noisy image. They weren't there in the program itself, but they are a result from video compression. The noise on the other hand, that is gone gone fairly quickly.

[rickb1988nl]

@edwardkmett I meant the big jpeg blocks created by the video compression you can see as it converges. I know that the path tracing causes noise, I have my own path tracer as well and I have regular discussions with Jacco about his tracer. But I can see how my post can be wrongly interpreted.

[edwardkmett]

@rickb1988nl the noise isn't from compression, it is the nature of path tracing, you accumulate a bunch of samples that converge towards the actual answer for each pixel. when the camera moves, he needs to resample.

[obdeniye]

Mark my words. What you are seeing here is the future of computer graphics. Not only is this realtime ray-tracing, but this is realtime ray-tracing *with* global illumination. The scenes are simple now, yes. But just wait 15 years.

[phoenix13nl]

What artifacts? If you mean the noise, I don't think it's the compression. Bidirectional path tracing requires an ridiculous ammount of samples, which wouldn't work very well in real time, I think this render takes just enough samples to produce a real time image, and when the camera is steady it takes enough samples to produce a pretty image.