Quote:
Originally Posted by luka
Hi Bojan,
Read the deep sky stacker page I have attached in my previous post. I am afraid you are not correct.
From the article:
...Of course when you subtract one dark frame to one light frame you will remove the hot pixels which may give the false impression that the dark subtraction did its job.
However, at the same time the subtraction doubled the noise of the calibrated light frame and ruined it thoroughly...
And ICNR would be exactly one dark subtracted from the light unless some other filtering is applied (which I am not aware of).
Remember, while the noise is random, it still follows the standard error analysis and it is added when two signals are added or subtracted. If you have 16+-4 and 9+-3 the result from the subtraction is 7+-7 (and not 7+-1 as many people think).
|
Luka, this statement in the article is misleading .
It is true, though, that hot pixels are removed.
Also (as I stated) the mean value of noise is also removed (or, shall we say, its DC component, or bias).
However, noise (which is superimposed on bias mean value and which is randomly distributed around it) will not double with subtraction of the dark frame.
If we are strict here, it will double the noise POWER (which is proportional to mean amplitude, squared) but that results in 1.4x (=SQR(2) higher noise amplitude, and not linearly, as in your example. Actually, even this statement is not quite right because noise distribution with CCD (and I think CMOS is the same) is binomial, and not Gaussian...
Now, during the stacking process, those noise peaks are averaged out (stacking adds the luminances of every pixel and then takes mean vaule from them or takes median - depends on settings) and the visible noise disappears very quickly.. only 3 - 4 frames are enough for first approximation.
I still think the benefit of not having to deal with dark frames later justifies having the INCR ON (for short exposures, of course..)