tag:blogger.com,1999:blog-33892275.post8687336283424511128..comments2014-07-18T22:45:33.564-04:00Comments on erratic semaphore: Intersecting Summation CompressionAnonymoushttp://www.blogger.com/profile/17427077588637523185noreply@blogger.comBlogger4125tag:blogger.com,1999:blog-33892275.post-73840651828054012122006-09-07T22:39:00.000-04:002006-09-07T22:39:00.000-04:00In an earlier version of this problem I imagined a...In an earlier version of this problem I imagined an algorithm that would encode especially high-chaos areas for easier reconstruction later. This is, essentially, a generalization of "chopping up" the image: if you have a big area with nothing but "white space", there's just no reason to subdivide it. Here the connection to subdivision (in more of a <a href="http://i33www.ira.uka.de/applets/mocca/html/noplugin/inhalt.html">graphics</a> sense, but still related to the quadtree sense) becomes apparent, and subdivision is isomorphic to wavelet compression (hence, compression in general), inpainting/filling-in, and the general pattern recognition/learning problem. So once you start thinking of subdividing it, it immediately turns into a <i>very</i> general problem (surface subdivision/wavelet compression/inpainting/pattern recognition), which already has plenty of researchers :) I'm more curious if a "pure" version of this problem can be solved than whether related techniques can successfully dissolve the problem.Kylehttps://www.blogger.com/profile/16969956393774965833noreply@blogger.comtag:blogger.com,1999:blog-33892275.post-53241131817097041352006-09-07T21:34:00.000-04:002006-09-07T21:34:00.000-04:00Have you tried chopping up large images into an Nx...Have you tried chopping up large images into an NxN grid of small images?Anonymousnoreply@blogger.comtag:blogger.com,1999:blog-33892275.post-43218295705341234982006-09-07T19:11:00.000-04:002006-09-07T19:11:00.000-04:00Kind of picross-ish, except with real instead of b...Kind of picross-ish, except with real instead of binary values. You're right about the AI techniques, I had a solution that worked on small images using genetic algorithms. For bigger problems... let me know if you find a solution.Kylehttps://www.blogger.com/profile/16969956393774965833noreply@blogger.comtag:blogger.com,1999:blog-33892275.post-63851978976377625352006-09-07T15:15:00.000-04:002006-09-07T15:15:00.000-04:00Sounds like a picross puzzle. The question really ...Sounds like a picross puzzle. The question really is, can an arbitrary image be made into a picross puzzle, and still be reconstructed?<br /><br />Somehow, I doubt it, but with some clever algorithms (possibly AI techniques?) it might be doable. When I get some time I'll try poking at this one, sounds very fun.Anonymousnoreply@blogger.com