SIGGRAPH Souvenirs

[ingenious]

Let me also contribute a little something to the new visuals section. Guess where this work in progress will be presented

The caustic on the red wall is jaggy due of the coarse tesselation of the glass.

representative.png

[Thammuz]

Awesome rendering, it's fantastically clean...can't spot a single noisy pixel! But please, do tell more...Method used (path, ppm, bidir), time it took to render, samples/px, materials (brdfs), lighting setup etc...and how did you get the inside of the glass ball so clean?...stuff like that usually takes ages to clean out (for traditional methods, pm/ppm/sppm handles is well though).

[ingenious]

...and how did you get the inside of the glass ball so clean?...stuff like that usually takes ages to clean out (for traditional methods, pm/ppm/sppm handles is well though).

If I used path tracing, you would ask how I got the inside of the ball so clean. If I used PPM, you would ask how I got the diffuse so clean

The answer is based on a simple insight
BDPT + PPM = BFF

The used algorithm is based on a combination of bidirectional path tracing and photon mapping into a unified framework that handles both diffuse and specular transport efficiently. It doesn't use the concept of density estimation at all, it's very elegant and therefore very practical. Details will follow soon!

This is a representative image required by Siggraph that took 5 hours on a Core i7 CPU at 1500x1200 resolution. Materials are nothing special - glass, phong, diffuse.

[admin]

There was a presentation during Eurographics 2011 where a similar technique was presented (in the sense that a hybrid approach combines strengths of different algorithms, each approach handles a number of surface interaction types): the authors used path tracing for the primary rays and speculars, then vpls for indirect lighting, and photons for caustics. They had some very good results, but no paper, it was discussed during a session in which several ideas where presented. Anyone remember the details?

[ingenious]

There was a presentation during Eurographics 2011 where a similar technique was presented (in the sense that a hybrid approach combines strengths of different algorithms, each approach handles a number of surface interaction types): the authors used path tracing for the primary rays and speculars, then vpls for indirect lighting, and photons for caustics. They had some very good results, but no paper, it was discussed during a session in which several ideas where presented. Anyone remember the details?

I did not attend EG 2011, but there's a CGF paper by Dammertz and Keller called "Progressive Point-Light Based Global Illumination" that does the same.