Denoising is another obvious application. You could perhaps do your decomposition at various points over the image surface and see if you can find some noise function that's present everywhere and that therefore comes from the sensor rather than the object. Again, there is a huge literature and it's a difficult problem. http://en.wikipedia.org/wiki/Noise_reduction On 28 February 2015 at 13:54, Bill Skaggs <weskaggs@xxxxxxxxx> wrote: > Decomposing a distribution into Gaussians is the essence of unblurring. A > good algorithm for doing that would of course be very useful, but there is > an enormous literature on the topic, and the most important fact about it > is that it is mathematically ill-posed. In other words, unless you add > extra constraints, tiny changes in the source distribution result in very > large changes in the output. (In image-processing terms, transforms of > that type tend to create large artifacts.) Unless the new method has some > way of handling that problem, it probably isn't going to be useful. > >> > > >> > > http://benpaulthurstonblog.blogspot.com/2015/02/ >> > supposing-you-have-process-that-reaches.html >> > > >> > > On Fri, Feb 27, 2015 at 3:56 PM, Ben Thurston < >> benpaulthurston@xxxxxxxxx >> > > >> > > wrote: >> > > >> > >> I developed this type of function that I feel is sort of like the >> > >> statistical analogue of the Fourrier series, it breaks a distribution >> up >> > >> into simple normal distribution components as the Fourrier series >> > breaks a >> > >> wave into simple sine wave components. I thought maybe there could be >> an >> > >> application for it in image processing but I don't know enough about >> > image >> > >> processing to figure out how it would apply... Anyone have any ideas? _______________________________________________ gimp-developer-list mailing list List address: gimp-developer-list@xxxxxxxxx List membership: https://mail.gnome.org/mailman/listinfo/gimp-developer-list List archives: https://mail.gnome.org/archives/gimp-developer-list
