jpgordon@well.sf.ca.us (Joshua Gordon) (09/14/90)
Hello! I'm developing a program that (among other things) will be implementing the algorithm on page 108 of "Science of Fractal Images" by Peitgen and Saupe. It's working fine; but I've come across something I don't understand. On pages 106 and 107 are a sequence of spectrally synthesized mountains. On page 109: "In the sequence...we show how the spectral synthesis method 'adds detail' by improving the spectral representation of the random fractal, i.e. by allowing more and more Fourier coefficients to be employed. The resolution...was N=64 but in the top image... only the first 4 coefficients were used. In the other pictures we allowed 16 and 64 non-zero coefficients..." Now, here's my problem. I'm not sure _where_ to pick the "first M coefficients.", and how to pass that information to the Fourier transform. Obviously, some of the coefficients need to be zeroed out, but which ones? The key to making it work right is satisfying the conjugate symmetry condition mentioned on page 109, but I'm still at a loss figuring out which are the "first" ones. It is not just the first M ones calculated; the algorithm as presented does a fairly optimal job of calculating the things. (Though there are some redundant calculations.) Could anyone help me with this? Fast? Help? Respond by mail (if it is interesting enough, I suppose you could post it here too!) to either the reply-to up above, or to uunet!autodesk!josh Thanks!