"Noise-aware" MLT

[tomb]

Stumbled upon this:

http://graphics.cs.uiuc.edu/~jared/impo ... gf2009.pdf

(it was published in september 2010, http://onlinelibrary.wiley.com/doi/10.1 ... x/abstract)

Seems like an interesting approach, i.e. guiding it through something more than Y - anyone else read this paper?

T

[jeanphi]

Hi,

Yes, I read it with great interest. I have a plan to introduce variance data to the film, the scheme would then be pretty easy to fit into the existing code.

Jeanphi

[tomb]

Hi,

Yes, I read it with great interest. I have a plan to introduce variance data to the film, the scheme would then be pretty easy to fit into the existing code.

Jeanphi

Sweet As with all new papers, I've learned to be very cautious about all claims to greatness - but it does seem like a very useful and fairly straight forward extension

T

[patrickwalz]

i really liked the idea of the importance shader -- i could easily see something like favoring specular paths to help caustic converge more quickly -- variance data would be really useful in the future should we implement pmcer, ect

[vimax]

That could really be very useful when you are over 10k s/px and some dark area just doesn't really clear up nicely. On the other hand I wonder how it behaves when it's rendering a noisy texture with a sub-pixel noise, wouldn't that take the interest of the algorithm all the time?

[viewon01]

Have you experiment it ? It sounds to be an interesting idea !