In particular that run natively on Linux.
Just for the record, you mean open source as in free app or app where you need to access the source code?– t3mujinDec 6, 2010 at 15:16
@t3mujin: let's assume Open Source as in opensource.org/osd.html– mattdmMar 23, 2011 at 3:06
There are several different plugins for gimp to remove noise.
Can you comment on the differences between them and their relative advantages and disadvantages? How do they compare to the state of the art in proprietary software?– mattdmDec 4, 2010 at 0:36
Addendum: that search result isn't particularly helpful, as it contains several discontinued projects, and links to pages that say something like "I have GIMP 2.6.6 (GIMP.app) running on Mac OS 10.4.11. I need a noise reduction plugin that will work for this version of GIMP".– mattdmDec 4, 2010 at 5:19
GREYC's Magic Image Converter (G'MIC) is a continuation of GREYCStoration. I'm not a fan of it, as I've found it to be very slow and there are loads of options to tweak. This means you could spend hours just tweaking the parameters to find an optimal set for a particular image.
My personal recommendation would be Wavelet Denoise. It's fast, has few parameters and generally produces excellent results. The results are not quite as good as the commercial packages, but they're close enough that you won't notice the difference unless you're pixel peeping. For best results, I use the YCrCb colour space and apply noise reduction to the Cr and Cb channel. Go very easy on the Y channel (the luminance channel), because if you over do it you'll end up with a very "plastic" looking image. I normally leave the Y channel alone as I like a little bit of grain in my images.
Yes, I have found wavelet denoise to work very well. Additionally, Ufraw has wavelet denoise built in.– labnutDec 6, 2010 at 17:58
I can tell you a very simple trick that you can do it with any photo editing tool.
1) Double your image size,
2) blur it a bit,
3) bring it back to the normal size.
1Depending on your 'blur it a bit' step, this isn't much different than doing a blur with a kernel that's half the size as the one you use in step 2. The only real difference is that you'll introduce aliasing artifacts from the resolution changes.– mmrDec 6, 2010 at 6:17
@mmr Yes there is a different on digital mathematical world. This is digital, so when you double it you actually create new digital points. And when you blur them - you can even add some extra small noise inside them...– AristosDec 6, 2010 at 13:57