Using Harris corners to analyze the picture is a cool idea too. Good job!
A cubism app could 1) detect edges 2) distort the image in a way that these edges follow splines 3) place things enclosed in edges at a random offset 4) amplify edges.
And besides, when art takes itself too seriously, with strict descriptions, it limits itself. I think Braque and Picasso would think that this is neat. So this is great.
The parent has a good point though, this looks nice and is interesting but wouldn't it misses the main tenet of cubism, which was to show multiple facets of objects simultaneously, so taking a face, you might see an eye and nose in profile, and also an eye from the front - it is like a simultaneous view of all sides of an object, not a distorted view of one side.
http://www.metaphorical.net/note/on/kubist (really old website, sorry it's kind of unusable :P)
I love that you can generate these patterns from images, the gradients are a killer feature!
I've made a tool that also generates Delaunay Triangulation patterns but lacks importing of an image. Check it out as well: http://msurguy.github.io/triangles/
 - http://bost.ocks.org/mike/algorithms/
The "Analyze image" uses JSFeat to apply a computer vision algorithm (feature point detection) to add some "interesting" points based on the image.
Then for each polygon, I take its points (center point or vertices or an arbitrary point) and find the corresponding pixel on the image. Then I extract the rgb color, and use that to fill the polygon (as solid color or gradient).
love code art.
jenn schiffer has some nice stuff @