Hmmm, it seems like he took some standard algorithms for deconvolution and made them into an application. Useful but not earth-shattering (although I have not looked at the source code). Depending on what kernel an image is convolved with, the inverse (deconvolution) may or may not be possible without loss of information. Even if the convolution by a particular kernel is invertible, it will usually be close to being singular (non-invertible) and therefore it will amplify noise in the image. This process is analogous to trying to undo low-pass filtering by amplifying and high-pass filtering. Any noise which is added in-between the low-pass filtering (which you cannot control since it is caused by the out-of-focus condition or the hand motion in this case) and the post-processing, will be amplified by the process. This goes for quantization and compression effects, too. If you already have some additional information about the image, you can make use of that to get a better de-convolved image, but only if you didn't already use that information in compressing the image (i.e., this method is likely to work much better on, say, raw or losslessly-compressed images than on JPG). Sean On Tue, Oct 23, 2012 at 7:41 AM, IVP wrote: > A software engineer has developed an app that can de-blur extremely fuzzy > images with impressive results. > > Called SmartDeblur, Vladimir Yuzhikov says the app attempts to solve 'one= of > the most interesting and important problems of image-processing' > > http://www.dailymail.co.uk/sciencetech/article-2221466/SmartDeblur-Downlo= ad-app-blurs-fuzzy-images-amazing-results.html > > with downloadable app > > http://yuzhikov.com/projects.html > > -- > http://www.piclist.com PIC/SX FAQ & list archive > View/change your membership options at > http://mailman.mit.edu/mailman/listinfo/piclist --=20 http://www.piclist.com PIC/SX FAQ & list archive View/change your membership options at http://mailman.mit.edu/mailman/listinfo/piclist .