I’m currently experementing/musing with the next phase of SSS. If I can get a volumetric image of the internals of an object I not only get that, but some other exciting things like cool fracture data for destructive physics. :spin:
That gets pretty cpu heavy in a hurry! Ideally, I’d like to use wavelets to store/calculate this stuff like this paper is hinting at: http://www.nlm.nih.gov/research/visible//vhpconf98/AUTHORS/PARK/SECTION2.HTM
Yet I don’t quite understand how a 3d wavelet is made. I’m guessing it’s a lot more complicated than 3 standard 1d wavelet results, mattrix multiplied.
As the author of the source code mentions, extending wavelet transform into higher dimensions is fairly straightforward. It is the same technique as doing a 2D or 3D Fourier transform. If you were to do a wavelet transform on an image (2D), you would first apply the 1D wavelet transform on all the rows individually (X dimension), and then you would apply the 1D wavelet transform on every columns of the previously transformed rows (Y dimension). To extend that in 3D, you just need to do do the 1D wavelet transform in the Z dimension.
Have you considered using existing wavelet packages. IIRC the R statistics program ( http://www.r-project.org/ ) has several wavelet packages. I used an R package called wavethresh to do denoising of experimental data. At the time I knew even less about R than now so I just called the relevant functions from python using the RPy interface. Perhaps you could do the same.