ICNR versus Darks comparison on the Southern Cross

[dugnsuz (Doug)]

Hello all...
This started off in another thread about digital thermometers!
I had considered putting together a darks library for use with my DSLR but some experienced imagers questioned the point of this.
Instead, I thought I would compare the use of dark frames taken over the course of an imaging session versus the same number of subs taken with the in-camera-noise-reduction (ICNR).

Details...
Canon 40D (modded)
Canon 70-200mm f4L lens
HEQ5 Pro mount

All images: 70mm, f4, iso1600, 16x5mins (80mins total)
All images (apart from final completed image) were processed minimally - stacked/combined in ImagesPlus then Levels (Auto button) in Photoshop.

1st set of images used ICNR (fullsize image and crop detail)
High Res...
http://s327.photobucket.com/albums/k...rrent=icnr.jpg


2nd set of images used dark frame subtraction in ImagesPlus. I took 3 images at the beginning of the session, 3 in the middle and 10 at the end. 16x5min dark frames in total - Median Combined in ImagesPlus.
By not using ICNR I managed to take a total of 39 light frames which was great. But, I've only used 16 in this example to match the number of ICNR subs.
High res...
http://s327.photobucket.com/albums/k...rent=noise.jpg

Final image is a stack of all the subs (55x5mins!!!)
High Res...
http://s327.photobucket.com/albums/k...urrent=10a.jpg

If you do a blink comparison of the 2 cropped pics, the dark frame subtraction looks better to me I think. There are some stacking/focus issues with the ICNR crop though, but regardless of that, I feel there is less red graininess to the black ares in the darks-only crop.

So, I'm sold on darks now - my imaging projects are now tending to stretch out over a number of nights and can be often broken up or have to be abandoned due to bad weather or time constraints etc.
Being able to pack in many subs in one session is a godsend.
Can't believe it's taken me this long to try a comparison like this!

Cheers
Doug

Pics are...
1.ICNR 2.ICNR Crop 3.DARKS 4.DARKS Crop 5.Final/Completed

[TrevorW]

A discernable difference Dougie, nice images by the way

[bojan]

That was the expected result.
Mathematical analysis shows that the S/N is better with darks subtraction.. I posted a link to webpage explaining why it is so couple of week ago.
However, even intuitively you can figure it out:
Any image has signal and noise.. subtracting Dark frame immediately after means adding additional noise to the frame (because subtracting noise has the same result as addition, sign does not matter here).
Stacking process is actually averaging, and because of that fact the noise is reduced. More frames, the better S/N ratio.
After that, subtracting noise stack (again averaged) frame removes permanent sensor faults from final image.

[mswhin63 (Malcolm)]

I agree, Darks from your images are better. I may use ICNR occasioanlly wil I get better at AP.

[dugnsuz (Doug)]

Thanks all for the comments.
Hope this comparison helps (visually) in answering the ICNR v's Darks questions that often crop up on the forum.
All the best
Doug

[Phil Hart]

great doug.. i love a good test!

there's no better way to improve your own imaging skills than prove what really matters. for somebody just getting started, ICNR gives a sound result while you learn all the other ropes. but once you get serious, dark subtraction is clearly the way to go.

once you add in the benefit of getting twice the number of subs in the same time 'under the sky', the difference is even greater.

your test shows something else that i've never really understood.. red pixels only make up one in four of the pixels in a DSLR yet the dark noise seems to be predominantly red. i have seen this in several different Canon DSLRs, before any processing (and it has nothing to do with modification since you can see it in 'dark' frames, and i've seen same thing on my unmodded camera too). Any ideas anybody?

[dugnsuz (Doug)]

Originally Posted by philiphart

your test shows something else that i've never really understood.. red pixels only make up one in four of the pixels in a DSLR yet the dark noise seems to be predominantly red.

Perhaps sensor heat "stimulates" all 4 pixels - but red predominantly, so we get that dirty brown/brick-red noise?

I'm keen to try out the darks on DSLR Ha images - that's where the strength lies for me as it takes stacks of 15 minute subs to get decent results for me, so the ICNR route was very time consuming to produce images like this...
http://s327.photobucket.com/albums/k...rrent=25-1.jpg
I think that took about 3-4 nights of imaging!!

[Tilt (Michael)]

Welcome to the fold Doug

I dumped ICNR in favour of taking dark frames a while back. Why? Because I was getting better results, simple as that. Normally I just take a bunch of light frames, then at the end of the run the cap will go on and a chain of darks will be taken. However, occasionally depending on the temp, I will pause the lights run to take darks, then take more a the end. Sinch!

Michael

[RB (Andrew)]

I stopped using ICNR ages ago and never looked back.
Also, I know it's a bit OTT but I try and take as many darks as I do lights.
Not very efficient, especially for a one night session, but I like the results I'm getting.

[Hagar (Doug)]

A few interesting things have come from this discussion. Clearly the dark subtraction is the winner to my eyes at least.
Dark frame subtraction allows the subtraction of any stray light in the imaging train which is not seen by ICNR be it just the smallest reflection inside optical tubes from an LED or just some light leak.

Great discussion and test.

[RobF (Rob)]

Once you do get a decent library together you will find/feel your nights are a lot more productive too. Watching paint dry is more fun than waiting for that wretched red light to go off and your DSLR to come back to life....