Here's another test image. This one isn't my photograph, nor is the subject my daughter.
But it is my processing — 250 circles this time.
There is an interesting effect where there is actually
more detail in the image than you can see by looking closely. I think this is because the lines of the circles interrupt your brain's image processing in interesting ways, preventing you from seeing the forest for the trees. A physicist friend of mine guesses that this is literally like the high-frequency noise on a short-wave radio channel — the voices are there just fine underneath, but you can't hear/understand them because of the squeal. For this reason, the blur induced by squinting really does something that just moving back or shrinking the image doesn't quite duplicate.
Compare to source image:
The project was inspired by this:
Genetic Programming: Evolution of Mona Lisa Roger Alsing Weblog
I realized that because he doesn't constrain the complexity of polygons used, what he ends up with is a very poorly color-quantized image. The
process is interesting, but the result isn't. In fact, as the complexity increases, it becomes more and more indistinguishable from something that could be done with a carefully-applied mosaic filter. (See
Genetic Gallery Roger Alsing Weblog)
I set out to focus on interesting end products instead — the circles are a part of that, for the same reason iambic pentameter is a good way to craft verse. (Although my program actually has some other interesting options I'm exploring.)
I started working before Alsing had source code available, and since I had gone in a very different direction before he published source, I haven't really even looked at that.