I've noticed a problem which, surprisingly, seems to affect all Ultra Fractal outside colouring formulae that use the running derivative approach from the Mandelbrot Set distance estimation algorithm (DEM/M) documented on page 196 of the Science of Fractal Images (SOFI).
See the three sample images below which were all produced using the Distance Estimator formula from Standard.ucl.
Render of the largest bud attached to the top of the main cardioid. The filaments on the left side (in red box) are much narrower than those on the right hand side (blue box). My understanding is that the filaments on both sides of the bud should be of similar width and should definitely not exhibit the dramatic differences shown in the image. A nicely balanced rendering is shown, for example, on page XII of the Beauty of Fractals.
Render of the largest minibrot to the North of the main cardioid. Some filaments to West and South are rendered too narrow.
Render of the area East of the largest minibrot on the real axis. The filament overlying the real axis is broken in parts.
Many deep zooms featuring visible minibrots exhibit similar imbalances in filament width around the minibrot. For images where the minibrots are too small to be seen and the main Mandelbrot Set body is not visible, filament rendering seems to be OK, i.e. filament width is broadly consistent across the image.
I'm seeing the same problem with distance estimator formulae that I've written myself. I also use the running derivative approach (without the overflow test, which as far as I'm aware is not often used and may only have been included as an optimisation).
The three sample images were all rendered using the one-pass linear drawing method. However, the same effects are visible for all three drawing methods, so the problem is definitely not a side effect of drawing method.
I'm fairly sure this is not a result of an issue with the DEM/M algorithm itself, as it has been successfully used in many programs. None of the DE images in the various books written by Peitgen and colleagues exhibits the problem. It's also absent in Fractal Works for Mac, which definitely uses the running derivative. See the two beautiful minibrot renderings in this post which show no sign of filament width irregularities:
-
http://www.fractalforums.com/mandelbrot-and-julia-set/how-are-bandw-images-from-'beauty-of-fractals'-created/msg4290/#msg4290Has anyone else observed this problem when working with distance estimation formulae based on the running derivative in Ultra Fractal? Are there any known workarounds or solutions?