jorkata wrote: ejmartin wrote:
SNR is scale dependent, because noise is scale dependent.
Yes.
And noise tests at 100% magnification tell you the percentage of noise per megapixel count.
This is a good measurement when cropping, for example.
The noise percentage does not change as you crop.
So, it can tell you, for example, how much noise will be there in a 7D image cropped to 40D dimensions.
The same at the same image scale as the 40D, if the 40D is cropped to the same angle of view.
To resolve this, all that‘s needed is to devise a formula that translates noise percentages into noise per print size and/or ppi.
With this formula we’ll be able to more accurately estimate noise.
You realize that such formula will ultimately be based on pixel-level noise, though, right?
The appropriate characterization of noise is its power spectrum as a function of line pairs per picture height. The only way that the pixel size enters this characterization comes from the cutoff on the spectrum at the Nyquist frequency, which will be lower for a larger pixel camera. The noise power at any given frequency in lph will be largely independent of pixel size.
As I mentioned, I don\'t understand the moaning about noise in images that larger pixel cameras can\'t generate, while the image that a larger pixel camera can generate, can also be generated from the smaller pixel camera, if desired.
To me, the issue is what is the best tool for a given use. The finer pixel cam provides the same image information at image scales the smaller pixel cam can generate, and additional information at image scales the larger pixel cam cannot generate. The same information where potential uses overlap, and more information for a wider array of potential uses.
Oct 11, 2009 at 07:25 PM
Previous versions of ejmartin's message #7627733 « Canon EOS 7D Master thread »
