I've generated some 2nd order SH light-probes from cube-maps in my scene and I was surprised how strong artifacts are around the back of the dominant lights. I switched to a Lambert source plus ambient and got a result just like in the image below. Stupid SH Tricks seems to be suggesting a convolution filter of some kind to filter out the ringing, but I'm clueless on what a Han function is or how to apply it. Little help?
[attachment=19354:ringing.jpg]
Not sure if this is relevant or would answer your question at all, but there was a talk at GDC a few years ago about using dealing with SH ringing by using something called needlets
http://gdcvault.com/play/1015310/Frames-Sparsity-and-Global-Illumination
Not sure if this is relevant or would answer your question at all, but there was a talk at GDC a few years ago about using dealing with SH ringing by using something called needlets
http://gdcvault.com/play/1015310/Frames-Sparsity-and-Global-Illumination
If I remember right this is microsofts fantastically complicated "Better than Spherical Harmonic basis spherical harmonic alike" that they used for Halo 4 and took a PHD in mathematics to come up with. I should probably brave going through it, but as far as I know its a lot more, and a lot more complicated, that just dealing with ringing
Not sure if this is relevant or would answer your question at all, but there was a talk at GDC a few years ago about using dealing with SH ringing by using something called needlets
http://gdcvault.com/play/1015310/Frames-Sparsity-and-Global-Illumination
If I remember right this is microsofts fantastically complicated "Better than Spherical Harmonic basis spherical harmonic alike" that they used for Halo 4 and took a PHD in mathematics to come up with. I should probably brave going through it, but as far as I know its a lot more, and a lot more complicated, that just dealing with ringing
I don't think it was used in Halo 4...if I recall correctly this was just something that seemed promising but ended up getting written off as a dead end for the time being. The GI basis for Halo Reach used SRBF's (which is actually a fairly popular area of research these days), so I would assume that this was also used for Halo 4.
I'm already struggling to understand what Stupid SH Tricks is suggesting, so I'm happy I don't have to deal with that paper. I've been looking for a sample of how to apply the windowing to make sure I'm at least thinking along the right lines, which is some kind of damping for smaller values on the lower base bands, but so far I didn't find very much.
I've attached actual images of my game this time.
At the front: Correct
[attachment=19369:sh1.jpg]
At the back: Too much green light showing on the other side.
[attachment=19370:sh2.jpg]
