Wavelets

[kruseborn]

I wondering if someone has implemented Environment Map with wavelets. And if they are willing to share some source code with me. I am sitting on project and have read a lot of papers, but having a hard time starting up.

I have search on google but its impossible to find some source code in this area. Only a lot of papers on the subject.

[Goz]

I'm trying to think HOW you could do environment mapping with wavelets ... but that maybe my limited understanding.

A year or so back I did texture compression and decompression of wavelet texture images.

Found performance and results to not be very good, tbh.

[kruseborn]

Here are some papers on environment maps with wavelets

http://www-graphics.stanford.edu/papers/allfreq/
http://graphics.stanford.edu/papers/allfreqmat/

[.oisyn]

I'm trying to think HOW you could do environment mapping with wavelets ... but that maybe my limited understanding.

Basically, spherical harmonics are analogous to Fourier transforms, but applied to a sphere. While FT (and SH) use harmonic oscillators (sine waves), the wavelet is simply a generic extension to FTs with other wave kernels than just the perfect sine waves. Naturally, you can apply this to spheres as well. So just think of it as SH but with different wave functions.

[Goz]

Makes sense :)

In which case, I DEFINITELY can't help ya ;)

[nomad421]

.oisyn

Generally, I agree with your characterization. However, the special property of wavelets is that they present a time-scale analysis of a signal, whereas the Fourier transform presents a frequency analysis of the signal. The special property of wavelets is that they are localized in the time frequency or space-frequency plane. It's this property that makes them powerful tools for representing signals (like lighting functions) which may contain sharp discontinuities. Wavelets can represent sharp changes well because they have local support; unlike fourier bases (or SH bases) which have global support.

kruseborn:

May I ask about the nature of your project? In general, I would classify these papers as Precomputed Radiance Transfer (PRT) rather than environment mapping; though this may be a difference in terminology. .oisyn's analogy to Spherical Harmonics is particularly apt in that these methods will have similar restrictions to lighting implemented using SH. However, an even greater restriction of these methods is that wavelets are not (in general) rotationally invariant like SH basis functions. There have been a few papers on approximating rotations of wavelet bases, but if you're looking to use this for lighting in an interactive, real-time environment (with moving objects), then perhaps this is the wrong approach.

Cheers,
Rob

[kruseborn]

Im doing my thesis on skin rendering, right now i am just testing different lighting implementations. Im doing my thesis at a game company so am not going to use wavelets at the end because they are to slow. But I think it would be nice to see the result of representing the environment map with wavelets and get the specular term with it. Not just the diffuse light you get with SH.