alexo Posted October 13, 2009 Share Posted October 13, 2009 There are some technologies from Microsoft Research that I'd love to see in Paint.NET (either natively or as plug-ins): [*:3vpyg5gh]GrabCut[*:3vpyg5gh]Patchworks for object removal.[*:3vpyg5gh]Blender for seamless object insertion[*:3vpyg5gh]Bayesian Color Constancy Is it feasible? Quote Link to comment Share on other sites More sharing options...
DarkShock Posted October 14, 2009 Share Posted October 14, 2009 Is it feasible? Maybe in ten years. Quote ---- Gallery | Sig Tutorial | deviantART | Sig Videos | PhotoBucket ----D E S T I N Y Link to comment Share on other sites More sharing options...
Neil Cassidy Posted October 15, 2009 Share Posted October 15, 2009 I mostly focused on #4, and I think it would be possible (but not easy) to turn the papers and source code into a working plug-in. You'd have to build the priors outside of Paint.NET. And note that the result would basically be an even-more-awesome version of Increase Local Contrast (and Local Contrast Enhancement). Quote Segment Image : Average Color (HSL) : Eigen Blur : []Cool, new forum! Link to comment Share on other sites More sharing options...
pyrochild Posted October 15, 2009 Share Posted October 15, 2009 Is Bayesian Color Constancy similar to viewtopic.php?f=16&t=28252 ? Quote ambigram signature by Kemaru [i write plugins and stuff] If you like a post, upvote it! Link to comment Share on other sites More sharing options...
Neil Cassidy Posted October 15, 2009 Share Posted October 15, 2009 In a sense. That corrects based on an actual color chart. The Bayesian algorithm corrects without needing a chart. It computes a posterior probability distribution for the color of the illuminant, based upon the observed intensities of the image pixels, a model of surface reflectance, and prior knowledge about what sort of color an illuminant might have. Then it estimates the true color of each pixel by "inverting" the illumination. I think the charts are just in the pictures to demonstrate its performance. The two plug-ins I mentioned also correct without a chart, but I imagine they use heuristics to do so. (edited for clarity) Quote Segment Image : Average Color (HSL) : Eigen Blur : []Cool, new forum! Link to comment Share on other sites More sharing options...
Neil Cassidy Posted October 15, 2009 Share Posted October 15, 2009 By the way, alexo, I thought I'd make a new post to thank you for asking about this! I had no idea that there was a generative model for pixel color until I read your post. If you're interested, the technique in #4 seems to go back to this paper: Bayesian Color Constancy with Non-Gaussian Models. The paper discusses several alternatives which may be simpler to implement in Paint.NET. Quote Segment Image : Average Color (HSL) : Eigen Blur : []Cool, new forum! Link to comment Share on other sites More sharing options...
alexo Posted October 15, 2009 Author Share Posted October 15, 2009 I mostly focused on #4, and I think it would be possible (but not easy) to turn the papers and source code into a working plug-in. I'm more interested in the other three I wonder if, due to Paint.NET being a showcase of Microsoft's .NET technology, the MS-Research people can be persuaded to contribute... Obviously I am not the person to approach them though. Quote Link to comment Share on other sites More sharing options...
Recommended Posts
Join the conversation
You can post now and register later. If you have an account, sign in now to post with your account.