PixInsight Forum (historical)
PixInsight => General => Off-topic => Topic started by: h0ughy on 2010 May 05 16:38:43
-
last night i managed to get over 4.5 hours of imaging done on Eta Carina. i know that when i finally finish stacking the shots and start to process that my field will be full of nebulousity. How do i go about doing DBE etc and processing the image? the other shots i have done have done hav been M83 and Omega Centauri and i have had background stars that i could colourbalance with? I have included a small version of one of the frames that i am currently stacking (87 x 3minutes)
-
DBE: You (almost) can't :D Try to find some very dark dust lines, and make use of symmetries (if possible) to cover the image. Use that procedure only for mild radial gradients, or almost linear ones. A good flat is irreplaceable.
Color balance: I use those dust lines too, but most times I don't perform any color balance based of the image information, since I already know the multiplication factors that apply to my canon sensor. Just set the black and white points as usual, and everything should go smooth :D
-
DBE: You (almost) can't :D Try to find some very dark dust lines, and make use of symmetries (if possible) to cover the image. Use that procedure only for mild radial gradients, or almost linear ones. A good flat is irreplaceable.
Color balance: I use those dust lines too, but most times I don't perform any color balance based of the image information, since I already know the multiplication factors that apply to my canon sensor. Just set the black and white points as usual, and everything should go smooth :D
apart from a night tinkering and not knowing what to set with the black point and the white point. i did take good flats - just that wont remove lightpollution.
thanks for the ideas though
-
Light pollution typically is a linear gradient. Try to determine the direction and then you may put some "fixed value" samples over the image to fix it. As I said before, use some dark dust patches as reference.
I set the black & white points using a statistical approach (automatic set). For the black point, 0.01%, and 0.002% for the white point. Then, just check the black points at each channel, so neither of them goes beyond the uprising curve, or has been left on the pure black.
-
Hi:
I think ABE woul'd do a good job. How about clasic PM expresion $T-med($T) for background calibration?
Color calibration is just another history.
Best Silvercup
-
Hi Silvercup
No, ABE will fail in this case, as your expression. Red will dominate, and this is fine. You just have to rely on your calibration frames, and use the dust lines to get rid of the light pollution (with DBE). For color balance, you may set the black point using the dust lines (since they are the darker features) and try to perform a balance based on the stars (like G2V or other better method).
-
Hi
You can get away with very few samples
Have a look here http://www.harrysastroshed.com/pixuser/DBEpart2.html (http://www.harrysastroshed.com/pixuser/DBEpart2.html)
Harry
-
Hi
You can get away with very few samples
Have a look here http://www.harrysastroshed.com/pixuser/DBEpart2.html (http://www.harrysastroshed.com/pixuser/DBEpart2.html)
Harry
thanks Harry
-
DBE without background patches is hard. Dark dust lanes aren't even the same as background as they're darker than the background. I sometimes create a registered Deep Sky Survey image that I use to find the dimmest areas of the nebula and use those for DBE samples. Still that won't work if there's no background.
For large fields of view I've found that gradients can be quite complex in the final image, especially for long duration exposures with perhaps a meridian flip thrown in. After all the gradient varies as you track the sky. At least in my case the LP is brighter as you get closer to the horizon.
To be fair though your image doesn't seem to have heavy gradients. Perhaps just color calibration can take care of it?
-
DBE without background patches is hard. Dark dust lanes aren't even the same as background as they're darker than the background. I sometimes create a registered Deep Sky Survey image that I use to find the dimmest areas of the nebula and use those for DBE samples. Still that won't work if there's no background.
For large fields of view I've found that gradients can be quite complex in the final image, especially for long duration exposures with perhaps a meridian flip thrown in. After all the gradient varies as you track the sky. At least in my case the LP is brighter as you get closer to the horizon.
To be fair though your image doesn't seem to have heavy gradients. Perhaps just color calibration can take care of it?
so how would you colour calibrate that image?
-
Ha, that's a good question :)
Clearly the stars are available as an average white reference but what to use as a background reference? If there's a small patch of dark I'd use that. I don't know what to do if there's not a single patch of dark. Someone else will have to suggest a better method than 'fiddle with the histogram till it looks nice'.
-
One method is to calibrate your filters (or bayer matrix) with another image. Find the right multiplicative factors that calibrates the image when it is linear. After that, all you have to do is apply the same stretching parameters to all of the channels. With this procedure, only minor touches are needed with curves, to get a more pleasant color, but you'll be close enough to the "real" ones.
-
Right but that assumes constant LP. In my case that would be a substantial assumption that probably wouldn't hold :)
-
LP is an additive effect. Once you substract it from the the image, you should get the right colors.
(Does you see the catch here? :D)
-
LP is additive yes. How does that contradict what I wrote? LP is not constant even if it is additive, certainly not for me. Imaging South gives me much more LP than imaging towards the Zenith. As I track an object LP will vary considerably per exposure. Not just the amount but also the angle of the gradient. The resulting stack has an 'interesting' LP component that can't be removed nor compensated for by filter calibration. I wish it was that simple :)
-
No no, what I meant was that once you calibrate the filter factors, then substract the LP, and after that all the colors should will be fine (and add a "background sky bias"... the median of the model works fine)... The problem, of course, is finding the LP gradient.
-
Hi:
I tried new background correction procedure with your image and this is the result:
(http://farm5.static.flickr.com/4055/4624882109_81341d8f00_o.jpg)
Has background correction "eaten" any nebulosity? What do you think?
Best, Silvercup.
-
Hi:
I tried new background correction procedure with your image and this is the result:
(http://farm5.static.flickr.com/4055/4624882109_81341d8f00_o.jpg)
Has background correction "eaten" any nebulosity? What do you think?
Best, Silvercup.
pm me your email address and i will email you the full res original jpeg to have a play with. looks promising what you have done
-
Hi:
Basically the procedure is similar to doing this:
Just extract background with AutomaticBackgroundExtractor with substraction in target image correction, but without applying to image.
Now you must apply this pixelmath expression with rescale result checked:
_MG_6698etaraw-_MG_6698etaraw_ABE_background/1.3
1.3 is a correction factor. It's easy to obtain this factor. Just do a preview on a zone you are relatively sure that is background. Apply a factor correction begining with 1. Inspect preview histogram. In your image you wil see that RGB channels don't match, the red is displaced. Increase factor till RGB channles match.
An image example:
(http://farm5.static.flickr.com/4064/4626283567_e65a74d77f_b.jpg)
That's all. I think this procedure do a great job.