Hi Andrew
Here's an explaination of the "Dark River Effect" an email from ImagesPlus Group.
Quote
'Several members of this forum have commented in the past on the dark
river effect, and asked Mike Unsold to look into addressing it with
IP. I have run into the effect in many of my photos. It's not easy
to correct for it. I thought I would share what I have learned, at
least what I think I have learned. Maybe you have some other
thoughts, test results, ideas, processing techniques, etc. Please
share what you know, so maybe we can get a good fix for it. Mike is
working on it now, so our collective input can help.
Here's what I know about the dark river effect on Canon 20Da's:
1. It is a darkened area that shows up in the sky background area on
longer (and deeper) astrophotos. It looks like a broad band across
the horizontal frame, running across the middle third of the frame.
2. The dark river shows up in both lights and darks. However the
effect in darks is NOT eliminated by subtracting the darks from the
lights. Dark calibration of the lights actually exaggerate the dark
river (you subtract a darkened area from a dark area, making it
darker).
2A. What is it? Or what causes it? It seems to be an area of the
chip that has less sensitivity to low level signals and noise. The
pixel levels are lower in the river by about 80 to 150 counts of a 16
bit full well. (That translates to less than a count on an 8 bit full
well). Treating it like a bias correction might work, but see note
10 below.
3. The river is stronger in longer exposures and higher ISO
settings. If you shoot 5 minutes or less, you probably won't see
it. On dim object astrophotos, the more you stretch the histogram to
see the dim nebula, the more the river shows up.
4. There is a temperature sensitivity, but I don't have enough
samples to fully characterize it. So far I believe the effect is
strongest at mid temps around 40 to 50°F. I see less at 30° and at
75°.
5. I have not heard anyone say they have the effect in any cameras
other than the Canon 20Da. Does it show up in modified or stock
20D's and Canon 350's? Or Nikon's?
6. You can see it if it's there using Images Plus>Color>Brightness
Levels and Curves>Digital Development and setting the Break Point to
500 or less.
7. The pattern of the dark river changes with exposure duration. On
my camera at 70°F, darks up to 20 minutes are generally a horizontal
band, or series of dark bands across the frame. At 30 minutes and
longer, the river turns into "Lakes" or blotchy areas across the
middle of the frame.
8. The areas of the river are not the same in darks and lights. I
had hoped that they were the same, so that darks could be used to
identify the river, without running into the subject of the image,
and be used to correct for the river. Unfortunately, this idea won't
work.
9. What works to correct for it? I have two methods, but neither is
perfect.
9A On astro images with a small subject and lots of uniform dark
sky, Use Jerry Lodriguss' method for flat fielding. See
http://www.astropix.com/HTML/J_DIGIT/VIGNET.HTM
9B Manually create an adjustment layer to see the river and create a
fill/mask layer (heavily blurred) with the inverse of the sky in the
river area. Play with the Fill and Opacity % controls to neutralize
the river, then turn off the adjustment layer and flatten the image.
This may require several iterations because the river is actually
several thinner bands, with different intensities. This method is
tedious and non exact, but allows trial and error if the subject of
the image is extensive.
10. What doesn't work? As mentioned above, you can't get rid of the
river by subtracting the darks, it isn't noise in the classical
sense. Correcting the darkened areas in freshly calibrated/un-
stretched images (or combined images) risks a sky background color
shift in the river if done before the color balance is adjusted for a
neutral sky.
11. What may work? I want to try this, but have to wait until the
next dark of the Moon. Take my light images of my selected target
for 10 minutes or more, than take one or several additional lights,
same setting, but move the scope off the target and into an area of
the sky that has few bright stars (these "river images" don't have to
be guided). Then use it ( de-starred and blurred) as a mask to
correct the river.
12. Mike Unsold has been asked to create an automated process in
ImagesPlus, and Mike has agreed to try. He may have an idea of how
to automate a corrective process, I hope so. I'm documenting what I
have learned to help him or others understand the problem and
contribute to a good solution.'
Hope that helps explain it a bit better.
Cheers
JohnG