How to beat the DSLR noise?

[Astrocava]

With these clear skies (at least in the center of Spain) I'm accumulating more objects that I can process. I wanted some fast processing before the deep one to show my friends but, lately, I hardly beat the noise. I'm asking all of you that use DSLR cameras for your methods to fight against the noise.

I use dithering and always stack more than 15 subframes, normally about 30. About 50 flats, 100 bias, and 50 darks.

Attached is the typical beginning. Only streched with STF and still linear.

Sergio

[Enzo De Bernardini]

My first advice is to try to use more light frames, like 60, 80 or more, depending the object. After that everything is easier  You can try with ACDNR using the built-in lightness mask.

Greetings,

Enzo.

[georg.viehoever]

What is the light exposure time? What camera? What ISO?
I usally take 60-120 lights of 30 seconds (bright objects), or 120 seconds (faint objects) at ISO800 with my Canon40D.

Georg

[Simon Hicks]

What paremeters are you using? ISO, exposure length, lens / telescope info, F#, filters etc.

You beat me to the question Georg 

[Astrocava]

30x600"
ISO 400. IDAS-LPS Filter. Scope and camera in my signature.

[astropixel]

This might be overkill, but 266 currently average 15 - 20 seconds.

[zvrastil]

I do not think it's overkill. I use exposures like 10 or 15 minutes quite normally with ISO 800 and modified Canon 300D. You need long exposures if you want to catch faint objects and details. You need some amount of collected photons to create signal recognizable from noise. And if your exposure is too short to create such signal, higher number of frames does not help to show it (simply because there is no signal in single frame). This is especially true for DSLRs, where quantum efficiency (probability of detection of coming photon) is something like 25-30% (so only one of three or four photons counts).

Try reading this text on DeepSkyStacker website http://deepskystacker.free.fr/english/theory.htm. Following quote is an excerpt from this page:

Are 100 x 1 minute and 10x10 minutes giving the same result?
Yes when considering the SNR but definitely No when considering the final result.
the difference between a 10 minutes exposure and a 1 minute exposure is that the SNR in the 10 minutes exposure is 3.16 higher than in 1 minute exposure.

Thus you will get the same SNR if you combine 10 light frames of 10 minutes or 100 light frames of 1 minute. However you will probably not have the same signal (the interesting part). Simply put you will only get a signal if your exposure is long enough to catch some photons on most of the light frames so that the signal is not considered as noise.

For example for a very faint nebula you might get a few photons every 10 minutes. If you are using 10 minutes exposures, you will have captured photons on each of your light frames and when combined the signal will be strong.
If you are using 1 minute exposures you will capture photons only for some of your light frames and when combined the photons will be considered as noise since they are not in most of the light frames.

regards, Zbynek

[astropixel]

It's interesting how opinion differs among authors. To some extent it also depends on other factors. At f/3.2 or even 2.8, under suburban skies, short exposures may work to advantage as far as sky gradients are concerned - at least l think this is what this post explains - http://www.samirkharusi.net/sub-exposures.html

[georg.viehoever]

...

Are 100 x 1 minute and 10x10 minutes giving the same result?
Yes when considering the SNR but definitely No when considering the final result.
the difference between a 10 minutes exposure and a 1 minute exposure is that the SNR in the 10 minutes exposure is 3.16 higher than in 1 minute exposure.
...

...

I think this explanation is not really correct. 100x1 minutes and 10x10 minutes capture the same amount of photons (i.e. signal). But having 100 pictures instead of 10 gives you much more (i.e. sqrt(10)) read noise, because every time a image is digitized the camera introduces read noise in its electronics. So the SNR of the stacked result is quite different.

There are other aspects that influence the choice of the optimal exposure strategy, such as the quality of your mount/guiding, brightness of the sky and the object of interest, number of planes passing the field of view etc..

Georg

[zvrastil]

Hi Georg,

you're probably right. The area of different sources of noise in digital imaging is something I always kind of ignored - I should take a look at it to understand it through.
Still I believe that 100x1 = 10x10 only if number of detected photons during 1 minute exposure is high enough - and for faint stuff, we really talk about several photons per minute. But read noise probably plays much bigger role in the result.
My personal experience is to use exposure as long as I can - then I'm able to reveal much fainter stuff in the final picture. Otherwise, what would be the reason to calculate sky-limiting exposure for instance?

Back to original question of Sergio. My suggestion is following: more exposures would definitely help, but I would not forsake length of the exposure to it. You can see that noise is worse in darker areas - where the SNR is low. This is normal. Some amount of noise in uncooled DSLR is inevitable (it's much better in winter BTW). When processing, keep the dark areas dark and use mask to protect them from stretching histogram and increasing saturation. Use inverted mask when applying ACDNR to protect areas with high SNR from damage.

Also, my personal opinion is that it's better to leave some small amount of noise in the image rather than trying to remove it completely. It gives image more realistic and perceptually sharper look. If you are too aggressive with denoising, image can get "washed" or "hand-painted" look. But of course, it strongly depends on quality of original data.

regards, Zbynek

[zvrastil]

It's interesting how opinion differs among authors. To some extent it also depends on other factors. At f/3.2 or even 2.8, under suburban skies, short exposures may work to advantage as far as sky gradients are concerned - at least l think this is what this post explains - http://www.samirkharusi.net/sub-exposures.html

I do not think the opinions are really different. You're trying to create best possible image under your current conditions. For my DSLR, it does not make sense to take longer exposures than about 20 minutes, because of the camera dark current. I'm usually limited by light pollution to 8-15 minutes (depending on place, weather and whether I'm using CLS filter or not). I'm not limited by mount and guiding - it is good enough to handle hour or more.

If you have bad sky, poor mount or no guiding, than these parameters are limiting your exposure. Then, it's better to use shorter exposures and still have decent pictures, than to acquire long exposures with strong signal but deformed stars due to poor guiding.

My point was, that using 600" exposure is not an overkill but an advantage, if your conditions allow it.

regards, Zbynek

[Astrocava]

Hi all!

These thread has diverged a bit from expected, but I'll try some ideas (playing with subs exp and number)

Will try too what Zbynek told, thanks  But when apply ACDNR? before or after stretching?

Have fun!

Sergio

EDIT:I've integrated different number of subs and run the noise evaluation script:

Code: [Select]

Autosave002(15 subs)

?R = 7.830e-004, N = 2726150 (33.91%)
?G = 9.117e-004, N = 4325353 (53.81%)
?B = 1.021e-003, N = 3414999 (42.48%)

Autosave003 (26 subs)

?R = 6.534e-004, N = 2855998 (35.53%)
?G = 7.398e-004, N = 4235010 (52.68%)
?B = 8.890e-004, N = 3496545 (43.50%)


[zvrastil]

But when apply ACDNR? before or after stretching?

As far as I understand it, it's designed to be run after stretching. That's how I use it. At least default edge protection thresholds seem to be set up for stretched images.
I would also suggest to watch this video tutorial: http://pixinsight.com/videos/DBE_M42_Example/en.html. I find it very instructive.

[Astrocava]

Well, I'm happy.  

I've registered and integrated the calibrated frames from DSS with PI and worked with masks to wash the noise.

Sergio