Let’s dig up an old project I worked on several years ago…. It’s something I’d love to get back to spending some more time on and being smart about it. The idea was inspired by Vincent van Gogh’s Starry Night. There’s so much chroma noise in the painting, but it forms patterns that make sense to us. I wondered if I could make a computer do something like that somehow? I spent a little time taking a totally naive stab at it. The basic idea is to take some sample pictures then paint another picture from them by statistically weighted random sampling. It just iterates along each column, looks at the colors painted so far next to it, and picks a color that has been observed next to them. Here are the input files I gave it (all put together into one for the purpose of this post):
And what it ends up producing is a little bit of entropy:
Obviously the program had no sense of the lines that need to flow through the images. The influence of the vertical scan direction is obvious in the output. It’s not enough to work with nearby colors alone. I’ve also worked on some edge detection software, so eventually I want to put the two together and see what I get.