Guess what!

a mixture between tower defense, shooter and rpg

hummel, what is that? o.o

fast fourier transform, left: original image, right: magnitude of the fft

...now I can set tree leaf normals for lod models too, after loading a wmb or creating a level (even for these ugly trees ):

@Hummel: Nice, although the missing large scale horizontal waves are suspicious. Are you doing that on the GPU?

@mk_1: the left image is actually the result.
@HeelX: jep, on the GPU. Concerning your missing waves: it's not tweaked at all, just the first results.

EDIT: ahh, you where searching for horizontal waves! Well, as you can see on the right site, the complex coefficients of the horizontal waves are blacked out, so this is intended. What do want with waves moving orthogonal to the wind direction?

I see! Can you please try it with the following image:
.

Have you tried to apply a filter in fourier domain and reconstruct the image? How fast is your implementation?

There are your magnitudes:




Of course, that was the first thing before I threw the phillips spectrum at it:

(right side shows only real component of the complex coeffients, not the magnitude (fault of stupid me two days ago))

Runs with 2.3 ms for a 512x512x64 image with FFT + iFFT. So it's nothing you want to do for a simple filter like a small/midrange blur.
EDIT: on a nVidia GTX 560 Ti

Have you tried screenspace effects? The interesting thing would be to make a lowpass filter dependent on the screen-size, so that the resulting blur is constant in time and quality (blurryness) for different screen resolutions.

Not much experimenting done yet, since I implemented it for wave generation in the first place and that's what I concentrate on for now. Also the fft requires a power of two input image, so you had to rescale your render target anyway. But as far as I know 3DMark11 uses ffts for wide range bloom.