Dear NG,
I'm looking for a fast way to produce 2d-noise images with 1/f or 1/f^2 spectrum. I currently generate the noise via inverse FFT, but since I need lots of data (~10^7 for a monte carlo simulation) it needs to be really fast. Does someone know a faster way than my approach?
- Dimensionality is between 20x20 and 100x100 - The spectrum doesn't need to be exactly pink/brown, an approximation is fine. - Implementation in either matlab or scientific python (LAPACK anyway)
Thanks a lot for hints, literature, algorithms! Mathias
-- http://mail.python.org/mailman/listinfo/python-list