Seeing as how Awake does much of its processing in Fourier space, would you consider making a transform tool? From spatial to Fourier mag/phase and back?
- Chad
I could, but is there a specific purpose for generating the frequency image? In general using the custom kernel input can achieve many effects that are beyond what the built in set was designed for.
There are some issues in generating the frequency image for Fusion though. The biggest problem is that 32-bit floats are often not precise enough and you start getting rounding error when you convert back and forth from the frequency domain (Internally, the plugin is working on 64-bit floats). The second issue is that they are slightly difficult to work with since the output image contains complex number pairs. There are few others regarding the size and boundary conditions, but if you know how to work in the frequency domain I can create conversion plugins easily enough.
Brian!
The idea for Frequency Blur is to reconstruct the image as it was originally, just blurred, but with very high fidelity. I don’t think I need that kind of precession. If a 32-bit output could “roughly” reconstruct the input again, that would be enough to make masks for correlation. I’m not sure how much precision would be needed to maybe do some band pass filtering or frequency suppression. It’s possible that the results could be used as masks if nothing else.
Not sure if it helps, but could you store the imaginary part be stored in another float image channel, like the NX, NY, and NZ? Or the background channels? BgR, BgG, BgB?
I’ll try the custom map, also. I’ve been meaning to, just haven’t played with it yet.
- Chad