Originally posted by pathdoc I'd be prepared to lose some resolution if I could cut down the noise a bit.
Interesting idea.
So I tried it approximately on my old Olympus which is noisy in low light by today's standards.
The camera was set to record the raw 3000 x4000 plus a low resolution 1200 by 1600 jpg
Here is the noisy raw (as tiff) increased in brightness
https://app.box.com/s/vdtet657ampi0qjiwdzps4bf22xjkgor
Here is the low res jpg, increased in brightness
https://app.box.com/s/qn64e4zpo9qpc64r46jupjxrs3i2f0vm
Undersampling reduces noise because it has to be followed by a low pass filter ( which may be the monitor itelf).
When an image is downsampled digitally before its resolution is limited, aliasing will occur.
I read that We don't always object to some aliasing due to the way the eyes work.
(compared to the ears listening to music with aliasing for example, which is objectionable.)
The bottom line is:
A photo is irreversible, so entropy must increase in any processing of it. That means a degradation of fidelity compared to the original sensor image, when any post processing is done.
Ref: John B Williams" Image Clarity" Chapter 5 gives a nice explanation with diagrams of Theory of Image Degradation, in terms of entropy and the degradation equation.
The composite spread function is sum-in-quadrature of the spread functions of each degradation. In my example the dominant spread function was by the undersampling.
There are probably more advanced ways to reduce the noise while minimizing the spread function, other than a crude undersampling.