×

INDI Library v2.0.6 is Released (02 Feb 2024)

Bi-monthly release with minor bug fixes and improvements

Dark Frame Subtraction in the Internal Guider

  • Posts: 1208
  • Thank you received: 559
I was recently thinking about the dark-frame subtraction option in the internal guider.
As I understand it, if checked, Ekos will capture a single dark frame exposed at the guider's selected exposure duration,
store that dark image, and later subtract that dark image from every future guider capture, performing the guider's
usual signal processing on the dark-subtracted guide image.

I'm posting because it seems to me that this might not be a good idea (i.e. subtracting ONE dark frame capture from the guide image).
I believe it likely introduces random noise into the guide image, and thus may cause more harm than good.
One could reduce the amount of added random noise, e.g. by capturing 5 or 10 dark frames, averaging them, then subtracting the averaged frame
instead of subtracting a single sample frame, but there will always be more random noise in the dark-subtracted guide frame, than the original guide frame.

I believe the reason to do dark subtraction, in spite of possibly increasing random noise, is to try and remove some of the fixed pattern noise
from the guide frame, as well as removing hot pixels. Hopefully detection schemes like SEP are not too sensitive to hot pixels, and less sensitive
to the fixed pattern noise, but perhaps "it wouldn't hurt to remove them and it may improve star detection in certain circumstances".

What do folks think?
- Is there a benefit to dark frame subtraction in the guider?
- Does it make sense to subtract a single dark frame?
- Would it be worth it to average N dark frames and subtract those from guide frames?

Hy
Last edit: 3 years 8 months ago by Hy Murveit.
3 years 8 months ago #57462

Please Log in or Create an account to join the conversation.

  • Posts: 1957
  • Thank you received: 420
In my recent experience, SEP is very sensitive to hot pixels. We have an ASI120mc-s which was poorly focused (stars were small balls) and guide calibration failed all the time because SEP chose a hot pixel. Once we better focused, SEP still occasionally chose a hot pixel and we still had to disable auto star selection and select one manually.


Wouter
3 years 8 months ago #57463

Please Log in or Create an account to join the conversation.

  • Posts: 969
  • Thank you received: 94
Hi Hy

I've been guiding using a zwo120mm and the new SEP multistar for several sessions and the dark frame subtraction either doesn't work (it keeps asking to take new frames) or when it does, makes no improvement to the guiding or the image produced from the guide camera.

@Hy: Doesn't SEP take into account background noise levels before star detection anyway?

I think that if this were to work properly, you'd need to make a master from several dark frames and subtract that instead. Otherwise a single dark frame is going to add <em>more</em> noise to the guide frame. This is how PHD2 does it; when setting up your guide profile, a dark library is made with say, 1s to 5s exposures. 5 dark frames are taken at each exposure and the master subtracted from the guide image, dependent upon which exposure is being used.

I'm happy with the SEP-multistar without the dark frame hassle. It just works.

Cheers,
Steve

** as an aside, dark frame subtraction doesn't seem to be logged.
Last edit: 3 years 8 months ago by alacant.
3 years 8 months ago #57464

Please Log in or Create an account to join the conversation.

  • Posts: 1208
  • Thank you received: 559
Wouter,

That is disappointing. sep_extract() is currently run with a 10-pixel minimum area for detecting stars--that 10 is not the area inside the HFR,
it is wider than that, I suppose meaning 10 pixels above the background level. See:
invent.kde.org/education/kstars/-/blob/m...sepdetector.cpp#L147
where the 4th arg, 10, is min_area:
invent.kde.org/education/kstars/-/blob/m...iewer/sep/sep.h#L219
Perhaps your system was fooled by some poorly calculated background level. In any event hopefully
Rob's upcoming SEP update, scheduled for 3.4.5 will have better optimized SEP parameters.

I did try to improve guide-star selection in the SEP MultiStar scheme. There it tries to select stars with the highest SNRs
(up to an SNR of 100). You might try to guide with SEP MultiStar to see if it improves things, if you aren't already.

If you have any guide images that are fooled by hot pixels, I'd love to check them out.

Hy
3 years 7 months ago #57492

Please Log in or Create an account to join the conversation.

Time to create page: 0.502 seconds