I'll start by saying I am neither a GIMP developer nor do I have any 
with algorithmic image manipulation. However, as a GIMP user I *REALLY* 
to make sure the GIMP development community is aware of the following 
since many
of you *DO* have the talent needed to handle this.

It seems there is a new mathematical technique available which can be 
used to
sharpen blurry or noisy images. From what I have just read it produces near
magical results. To quote from the article regarding the improvement in the
images: "It was as if you gave me the first three digits of a 10-digit bank
account number — and then I was able to guess the next seven."

I am also not a mathematician, but I can sort of understand how the 
works: It modifies the images repeatedly and on finer and finer scales 
always towards lower image complexity thus eliminating much blurriness 
and noise.
In point of fact almost all real-life subjects have edges and other 
The real world is mostly not blurred. This "compressed sensing" 
algorithm somehow
implicitly "understands" (note the quotes) that blur and noise is not 
what the
real world is about and it is able to supply missing data to correct 
images from
the real world.

Here is a pointer to a recent article on the subject published in Wired 
Please read it.


As GIMP user who looses a lot of images due to them being underexposed 
or out of
focus, I'd *REALLY* love to see a "compressed sensing" image sharpener 
plugin for
GIMP. I think many other GIMP users would like it a lot too.

Thanks for reading,
Jeff Barry
Acton, MA, USA.

Gimp-developer mailing list

Reply via email to