Interesting stuff. This is essentially a coding/compression problem. My advisor at UCLA helped pioneer the computer vision analoge for this. His work tackled en

Interesting stuff. This is essentially a coding/compression problem. My advisor at UCLA helped pioneer the computer vision analoge for this. His work tackled encoding textures (or areas of little information) with markov random fields and areas of high information (edges, features) with graphs. http://www.stat.ucla.edu/~sczhu/publication.html http://www.stat.ucla.edu/~sczhu/publication.html