PixInsight Forum (historical)
PixInsight => General => Off-topic => Topic started by: h0ughy on 2009 November 16 15:05:55
-
why is it that the "noise" of a DSLR always seems to be like a diagonal brushstroke? Its not just my camera either - i have asked a few other guys who use a DSLR
-
Most artifacts like that actually come from left over image artifacts being 'drawn' by the per-image drift in the stack. So there are a few causes:
- DSLRs create pretty 'rough' images
- calibrating these images exactly is hard, darks need to match your lights very accurately in temperature and exposures
- per image drift causes left over artifacts to draw streaks
Hard to tell from just an image what's causing it in your case but maybe the little list above gives you something to look into. Also read this:
http://deepskystacker.wikispaces.com/Measuring+differential+flexure
-
in this case the image this sample was taken from was stacked in deep sky stacker. 27 lights with darks, flats and bias done in deep sky stacker. I also get it in Images Plus.
Thanks for the link i will have a good read
-
ok i read that but i do not think it is explaining what is going on
-
I think that it is just dark current or electronic bias that has not been properly eliminated. Have you been taking the darks at the same temperature and settings as the lights?
-
Hi all,
Here is my take on the problem ...
Are you capturing RAW subs from the DSLR?
My firm belief, now, is that you CANNOT 'calibrate' images from a DSLR if you have saved these in any mode OTHER than RAW. This is because you will have NO KNOWLEDGE of the 'settings' used by the camera's 'internal' deBayer algorithm. The software could, understandably, use a totally different 'white point' for Darks than for Lights, for example. This would mean that you would NOT be able to use those Darks in your post-acquisition calibration steps.
If you are capturing RAW subs, then your next challenge is adequate temperature correlation between Lights and Darks - almost impossible in a camera never designed for temperature stability - far less temperature control.
The best you can hope for here is that the temperatures of any two frames exposed 'consecutively', are going to be relatively 'similar'. Some DSLRs (like my Olympus e500) take advantage of this fact and will automagically take a second 'Dark' exposure immediately after the Light sub, and then do the dark-subtraction BEFORE (internally) converting to RGB (if this is what you asked for) whever you have the "Noise Reduction" feature enabled.
But even this is far from ideal because, as we all know, the usefulness of a Dark Frame can be enhanced by 'statistical averaging' of 'many' individual darks. But you cannot (usefully) do this because you do NOT have the ability to 'temperature match' your Darks.
The only workaround I can come up with is to collect Lights and Darks (or Flats and FlatDarks) in a "Dark-Light-Dark-Light-Dark" sequence, and to then, manually, 'average' each pair of Darks surrounding each Light, before then using THAT, temporary, MasterDark to correct the associated Light.
Sure, it is a long-winded process, but it is the penalty that has to he paid for using a non-ideal DSLR instead of a dedicated astro-cam.
And, remember, you will have to repeat the D-L-D-L-D sequence for your Flats as well, an then will have to combine your calibrated Flats to give you a MasterFlat, which then needs to be applied to each Dark-calibrated Light.
Only then can you consider deBayering your calibrated Lights, and only thereafter can these be aligned and stacked.
There is nothing 'difficult' about any of these steps - you just have to be meticulous in the control of your data. It wouldn't take much for you to make mistakes, and I don't know of any software that can automate ANY of these steps - except, maybe, one of the 'scriptable' image acqusition programs that may allow the DLDLDLDLD.... sequence to be automated.
Does anybody agree with my thoughts ......
Cheers,
-
ok i read that but i do not think it is explaining what is going on
I didn't say it would. It *will* explain if you have drift or not. That is crucial. You can't fix a problem until you've properly diagnosed it.
-
Hi,
I understand that Dave Halliday http://pixinsight.com/forum/index.php?topic=1519.0 had a similar problem, and I understand that he got it fixed: "...by playing around in the "Kappa" settings in DSS I can eliminate 95% of the streaking".
I have never seen it in my Canon EOS40D images, even with my messy mount that has quite a lot of PEC and is unguided. I almost always use the "Recommended Setting" button in DSS to select the best settings, and those usually include Kappa-Sigma clipping. So I just don't know what I would see if I used plain averaging.
Kind regards,
Georg
-
Do you dither your subs?
Max
-
Hi,
yes, KS stacking can remove some of these artifacts but only if you have enough subs and only if you have movement between your frames. Keep in mind though that KS is not voodoo magic. Once the outliers have been rejected there are fewer pixels left to average so the SNR doesn't increase as much as you'd expect given the number of frames. This is OK when KS is used to eliminate hot pixels and satellite trails because a few pixels with more noise won't be a problem. When almost all background pixels have outliers due to bad calibration then the SNR of the background goes down.
I would follow these steps:
- measure drift (using the method I posted)
- if there is drift, eliminate it
- re-examine your calibration process and see if it can be improved to clean up the subs
- start imaging with dithering
- see if dithered images look better. They may because dithered streaks don't show up as much as diagonal ones.
- experiment with KS stacking
- take lots and lots of subs
- consider how serious you are about this 'hoby' and see if you want to upgrade to a cooled camera like a QHY8, Orion Starshoot, Parsec, ST8300 etc. etc.
I know the last step is not what you want to read but ultimately it'll boil down to that. At some point you reach the limits of what can be achieved with your equipment and then you decide if you want to move beyond that point. Unfortunately that's a journey that never seems to end, not even after buying a cooled camera :)
-
Well i have seen this before in my Dsi 1 colour camera ,and looking back i now know that my dark frames were not right, also if i had a stack of 20 lights i tended to think more the better, for my image, but really you have to be honest when looking at your images and throw out ones that are poor in quality,ie elongated stars,sky glow problems,and if after you have dark subtracted some are are not as clean, throw them out or redo darks. 10 images that are clean and tight will give a better chance of a nice image at the end.
Your Camera will take great images just takes time.
Alex
-
The image i posted was a crop from a friends Hutech modded 450D on a losmandy G11 f4 8". We were discussing the results and i commented on how my shots have always been like that - I have a cooled and modded 350D and a modded 20D. I noted that the darks in the 350D were far less "noisy" and i had more control over it. With the canon 20D i didn’t. Taking the temperature into account you would see more noisy pixels in the darks and in the lights. but it is only after they were stacked in DSS did the streaking become very apparent. I also noted that it wasn’t that prominent when processed and stacked in Images Plus. I am just wondering if it is the software introducing this feature. I have included the original drak frame where the correlating section of the image was shown - no streaks just dots
I would be interested in seeing some other results, and I will dig our some other data to compare the cameras used on the same night
-
The streaks came with the registration process. Your images do not exactly match, so a line of unfixed dark current is drawn. This may be solved in some way with a rejection algorithm during the integration, and/or a better match of the darks.
Well, this is the most likely cause... if it is something else, then I have no clue :D
-
these were taken from a image run with my canon350D which was cooled to -3. the guiding was out slightly in one axis (not sure why) but it doesnt exhibit as much noise as the uncooled camera.
-
Those four example subs that your posted - what was the exposure time for each sub?
Also, presumably you are using the Bias frames to calibrate the Flats (as opposed to FlatDark frames)?
Did you deBayer inside the camera (in which case, did you save as TIFF or JPEG), or did you save as RAW, and thus relied on an external deBayer routine?
Cheers,
-
Those four example subs that your posted - what was the exposure time for each sub?
Also, presumably you are using the Bias frames to calibrate the Flats (as opposed to FlatDark frames)?
Did you deBayer inside the camera (in which case, did you save as TIFF or JPEG), or did you save as RAW, and thus relied on an external deBayer routine?
Cheers,
they were taken straight from the raws - no conversion or enhancements (they were opened in CS4 just so i could layer the lot, crop a portion then save each layer seperately) then i saved for web because of the 128kb limit. the bias was a dark bias - no dark flat bias's - never done any of those. all taken iso 800 light wass 300 seconds, same for dark, bias was 1/4000 and the flat was about 1/250 i think?
-
So, I assume that you are using your 1/4000 Bias as a calibration frame for your Flat? And I also assume that you are taking the Bias frame as if it was a Dark, with the lens cap ON?
Personally, I would rather IGNORE the whole concept of Bias frames. They are just NOT necessary if you have the ability to control exposure time (which you do) and CCD temperature (which you can).
Instead of Bias frames (at 1/4000 sec) take FlatDarks (at 1/250s, or the same as your Flats).
Remember that a Flat and a FlatDark both contain (effectively) exactly the SAME 'bias' information and that, when you DarkSubtract the FlatDark from the Flat, you will successfully eliminate ALL traces of the Bias frame altogether (well, that's the theory, anyway).
And, exactly the same goes for your Light and (Light)Dark frames - dark calibration gets rid of the Bias information, as well as removing the Dark signal from your Light.
Does that make sense?
It shouldn't really affect the time taken for you to acquire the data. You will need to take, say, 20 FlatDarks at (say) 1/250" (whereas you might have been taking 20 BiasOffsets at 1/4000"). Overall, this is going to 'cost' you an extra 20 x 1/3750" - and I am sure that you will be able to spare that extra 5mS ;D
Whether this eliminates the 'striping' or not is another matter altogether >:(
Cheers,
-
I was just quickly skimming over this...
I am the guy who was all "in a dither" about trying to remove these streaks...
http://www.flickr.com/photos/daveh56/3868338463/sizes/o/in/set-72157622188544040/
I have found the use of "Kappa sigma MEDIAN" (in DSS) with a kappa of around 1.5,and 5-10 iterations works,when used with darks AND lites...
Mind you it takes like 30 minutes to stack...but I iron some shirts :laugh:
I am talking here about images taken with a Canon D1000...
I STILL have not figured out how to (easily) set up a dithering system,while guiding with PHD... ???
I think this WOULD also solve the problem,and improve my ability to use Drizzle in DSS...
I think dither is a secret that only the pro's know how to do...and they are not sharing !!
Dave