My second try to resume an image with Indigo was to test it's ability to take different values from the first render (actually my very first try didn't work at all, I was using the command uncorrectly).
More theory:
It seems logical to make evoluate the noise from very low frequency to high frequency, and max change from a rather high to a very small value, but effectively I had contradictory results. I didn't push the tests further enough to decide wich approach was more efficient.
But the very first step now is to define clearly what <max_change> and <large_mutation_prob> are exactly, and how they behave (linearly or not). Basically, I think that:
<max_change> concerns colors, how drastically they can be different from one sample to the next,
<large_mutation_prob> is the scale of the noise. Low values are low frequency noise AFAIK. Try very low values, you should even see clearly the stochastic path. Another example:
Do someone know exactly what these to guys are after ? Er... Ono maybe ?
If a plugin is made to manage these two, it should make them evoluate permanently from a start value to an end value (defined by user if not automatically (after analysys of the best response for start values)). If we have to resume by hand (or script) I think we should use "harmonics" of the start values for successive resuming.
... ok, I think I've said pretty much all I could...
If I remember correctly this is only the first iteration. You can see that large parts of the image are covered, while others were untouched.
Misc tests I did with noise settings (same render time for all images):
This is a 50% gray plane (1mn render maybe)Got to go, but you see there is a lot to do !