|
Here's what we do:
We look very closely at the histogram - the graph of the number of pixels with brightness 0, the number of pixels with brightness 1, etc, etc, etc.
If there are no pixels at all with a brightness less than say 10 out of 255 (8 bit histogram), or say 2750 out of 65535 (16 bit histogram), that means that there is NO INFORMATION in the image with a brightness less than 10 out of 255.
That means that we can set the black point of the image to 10 out of 255 (or 2750 out of 65535) with no loss of information. The result is a great increase in contrast.
For an RGB image, we do this on a per-channel basis.
It is crucial to do this process (what we call setting the foothill of the histogram to zero) BEFORE either applying a nonlinear stretch or fiddling with colour balance and saturation.
What causes that background wash of 10 out of 255 counts ?
The most obvious is dark current, and also the A/D converter's pedestal setting. None of that is of astronomical significance - it is just rubbish. But we normally get rid of that with carefully taken dark frames.
There is still a huge amount of rubbishy no-interest glow that is not due to subject matter, including:
(a) Light pollution
(b) Air glow, due to recombination of O and N free radicals formed by sunlight during the previous day.
(c) Moonlight
(d) Diffuse scattering of starlight by dust in the sky
The histogram trick will get rid of all that, with NO LOSS OF ACTUAL MEANINGFUL ASTRONOMICAL DATA.
Now you can see why it is important to set the zero point before fiddling with nonlinear curves, colour balance, or saturation. Doing those steps first will just let your moon glow, air pollution, etc overwhelm both the colour and the contrast of the image.
It is especially important to do this for narrowband images. Otherwise, one can produce very pretty images where the background seems awash with lovely blue OIII emission, but it is just artifact from the above causes.
Finally, there can be genuine structureless glow from the background to the object you are photographing - for example from unresolved stars, or from totally featureless nebulosity.
Consider the analogy of photographing a scene in fog. If what you want to show is the object against a background of fog, and the fact that it is foggy is important to you, then show it. If what you want to show is the object itself, and the fog is just a nuisance, then use a technique similar to what we've described to remove at least the totally structureless part of the fog.
However, we think that 99.995% of the background fog in an astronomical image is indeed dark current, light pollution, moon glow, and air glow, so it is a big mistake to try preserving that at the expense of your astronomical target.
Hope that helps.
Best,
MnT
|
)
