A distance estimator image using a new algorithm to determine the grey scale used. Pixels that don't contain set boundary are
black. Pixels that do are a shade of grey depending on the distance estimate at the sampled point, as follows: the distance
estimate in pixels is a value between 0 and 1 for pixels that contain boundary; that value's nth root is taken and then used to
determine the grey shade, from black at 1 to white at 0. If the boundary is dense in the pixel, the distance estimate is probably
close to 0 and the pixel ends up close to white. If the boundary is sparse in the pixel, the distance estimate is probably higher
and the pixel ends up a darker grey.
This seems to give superior results to my other method for some fractals. In this case, the eighth root was taken; if the root
parameter is set to 1 (so it just linearly shades from black to white as the distance estimate drops from 1 to 0) the image loses
much detail of the little scepters and around the minibrot.
This tinkering was motivated by a desire to generate a coloring algorithm that would do justice to all the frames of a zoom
animation without much difficulty. The ideal is for it to be "set it and forget it", which requires that at high roots the
zoomed-out M-set's filaments don't fade to black while details still don't "overexpose" in convoluted regions close to the M-set.
It's not quite perfect; for some animations the root would have to be gradually increased as the magnification raised, or perhaps
even varied in a more complex fashion. I may try further variations, such as dividing the distance range 0-1 into two segments
and using different roots for each, a low one for distances near 1 and a higher one for distances near 0. This would deal better
with animations, and with zooms of spike minibrots that wind up having "whiskers" as well as dense structures like seahorses. The
real bugbears though are images of minibrot spikes and similar environments, where the "whiskers" have seahorses and the like
attached. Getting DE to bring out the whiskers is nontrivial in these cases. For that, I need to consider mappings such as
.
This image was antialiased the standard way, but given the way DE pixel distances are calculated, it might work better to
generate a larger, unantialiased image (e.g. 8192x6144) and downsample it.
Freely redistributable and usable subject to the Creative Commons Attribution license, version 3.0.
Detailed statistics:
Name: Weave
Date: January 3, 2010
Fractal: Mandelbrot
Location: Triple Spiral Valley of top bud of top bud of M-set.
Depth: Shallow (15 decimals)
Min Iterations: 4771
Max Iterations: 1,131,813
Layers: 1
Anti-aliasing: 3x3, threshold 0.10, depth 2
Preparation time: 20 minutes
Calculation time: 4 hours (2.5GHz dual-core E5200)