For anyone wanting to give this a shot, I’ve managed to reproduce very similar drawings by using the following algorithm:
* divide the circle into x equal parts
* pick a constant stroke length
* pick a point that’s got the most black. This is your first “current point”
* then for each current point:
** try the x equally spaced directions from the current point
** pick the one with most black in the image
** subtract the black from the copy of the image in memory
** set the new current point
** repeat for the chosen number of paths
The result can produce some lovely-looking results, but it’s very random – very dependent on the stroke length, number of angles you try, and so on. The randomness is part of the charm.
In the Woolf image, part of the charm comes from the partial exploration – the line at the bottom follows one of the dress creases in the original image.
This is not something which will produce reliable results – it’s more something to experiment with.
While the Axidraw algorithm *may* differ from this, given the similarity of the results I’ve received to the image they supply, I’m reasonably confident that this is similar to what they’re doing.