Help with comet integration subtraction?

sforman

sforman

Well-known member
Hi all,
I imaged Tuttle a few weeks ago.  It's not a terribly interesting comet in a particularly boring part of the sky, but we had a window of clear skies in Northern California, it was up, and I had a couple of hours to kill. 

I have gotten as far as creating star aligned masters, and a comet aligned stacked master.  But when I try to subtract the comet image from the star aligned subs, I can't get it to subtract.  I'm wondering if it's something I'm wrong when I try to subtract, or if my data is just bad.  To do it, I control click all the comet star aligned files, identify the starting and ending time for the comet x,y coordinates using the first and last sub, select the comet integrated as the operand image, select "image integration", and "comet aligned", and I adjust the "reject low" to just above the comet background.  Still I can't get it to work, or at most, only one or two of the subs will reject the comet.

I am wondering if anybody can look at the data set? 
Here is the comet integrated master with the stars rejected
https://drive.google.com/file/d/0B6nhQcbEB-XUVUlHcmVKLVNNdmc/view?usp=sharing

Here is the lum star aligned, calibrated data set (15 subs).  I didn't upload the other colors. 
https://drive.google.com/file/d/0B6nhQcbEB-XUcEhLUFJuRVRiTWs/view?usp=sharing

Normally I'd just shrug my shoulders and move on to the next project, but for some reason this is seriously bugging me.  I hate it when I don't know why things don't work.  I've spent way too much time on this already--you know how it is.

Thanks in advance for any advice.
Stuart





 
I'm having the same issue. I can only get it to subtract a very small amount of the tail. I am working on C/2020 F3 NEOWISE. It is very bright, so I am wondering if that's part of the problem. I've tried adjusting the high and low sliders to every extreme and in between. Also have tried checking and unchecking "Normalize" with no success.
 
I'm having this same problem... I cannot get the subtract to, well, subtract. It reduces the comet a little but not substantially. I shot some data on Saturday night and 4 days later I'm still wrestling with it...
 
Typically the problem is that the linear fit is not valid between your combined (comet aligned) image and the individual frames that are being subtracted from. The reason that linear fit is not valid is usually due to poor data, or poorly calibrated data with large gradients (or a number of other vagaries). Usually what happens is that within a set of frames that are subtracted... you will find a small subset for which the linear fit is close or good..but the other frames diverge from this result and the subtraction becomes poorer.

I have now seen three examples of this effect. In all cases, the individuals either took new calibration data- or applied the calibrations more carefully so that the results were less noisy and flatted better. Then the subtractions were within expectations.

The goodness of the linear fit is really all the matters- and this is affected by large variations from the average from frame to frame.

-adam
 
ngc1535 said:
Typically the problem is that the linear fit is not valid between your combined (comet aligned) image and the individual frames that are being subtracted from. The reason that linear fit is not valid is usually due to poor data, or poorly calibrated data with large gradients (or a number of other vagaries). Usually what happens is that within a set of frames that are subtracted... you will find a small subset for which the linear fit is close or good..but the other frames diverge from this result and the subtraction becomes poorer.

I have now seen three examples of this effect. In all cases, the individuals either took new calibration data- or applied the calibrations more carefully so that the results were less noisy and flatted better. Then the subtractions were within expectations.

The goodness of the linear fit is really all the matters- and this is affected by large variations from the average from frame to frame.

-adam
Click to expand...
Could this "linear fit" issue occur more readily because of skyglow (and maybe low level thin clouds), given the position of NEOWISE etc? I shot a healthy number of subs, then darks and flats when I got home (without changing the imaging train) and used Batch PreProcess for the initial calibration and alignment. Not sure what I could do "more carefully" in that case (but more than happy to learn!). This was all OSC not mono, too.
 
Yes. Variations from clouds causing gradients in the images can certainly make a perfect linear fit difficult.
Please do not take my "more carefully" out of context. I was certain to add many "or's" in my statement.
I have not tried this- but it could be that an application of DBE on the individual subframes to remove gross gradients might help.
This is uncharted territory for me though.
-adam
 
ngc1535 said:
Yes. Variations from clouds causing gradients in the images can certainly make a perfect linear fit difficult.
Please do not take my "more carefully" out of context. I was certain to add many "or's" in my statement.
I have not tried this- but it could be that an application of DBE on the individual subframes to remove gross gradients might help.
This is uncharted territory for me though.
-adam
Click to expand...
Oh, I wasn't offended by the "more carefully" bit... I'm fairly new to all of this kind of image processing (but vastly not-new to terrestrial image processing), so it's all a learning curve in PI.

Is there, perhaps, a way to subtract images without using the Comet Align process? Something more controllable? I tried pixel math to subtract one from the other and it worked but looked... cranky, to say the least. And, thanks for your help!
 
You can do the linear fit manually... but there is no difference between what Comet Align is doing and just using the LF process. The benefit of the Comet Align process is that it shifts the averaged comet image (the operand) with respect to each subframe (in addition to the linear fit). I don't think there is a more elegant way... Comet Align is doing the simplest operation I think. You *can* manualyl force Linear Fit to do a better job by selecting an area of an image that is free from gradients and has quality signal using a preview. But this would be quite laborious. You would get the linear fit results and then with pixel math apply the results to the subframe based on the result from the preview.
 
ngc1535 said:
You can do the linear fit manually... but there is no difference between what Comet Align is doing and just using the LF process. The benefit of the Comet Align process is that it shifts the averaged comet image (the operand) with respect to each subframe (in addition to the linear fit). I don't think there is a more elegant way... Comet Align is doing the simplest operation I think. You *can* manualyl force Linear Fit to do a better job by selecting an area of an image that is free from gradients and has quality signal using a preview. But this would be quite laborious. You would get the linear fit results and then with pixel math apply the results to the subframe based on the result from the preview.
Click to expand...
Thanks for taking the time, Adam... I hope others find it useful. Now... back to the drawing board ;-)
 
Good day,
I am having a similar problem/
after star aligning the registered frames, I did the star alignment based on those frames with reference image one of the registered ones.
in the resulting frames a big black area is seen beside the C-2020F3 comet.
attach few frames to show the problem.
any hints?
I ave try all the setting in the star align that I can think about with no joy....

Blink00004.png
Blink00010.png

Blink00020.png

Blink00025.png
 

Attachments

  • Blink00014.png
    Blink00014.png
    156.8 KB · Views: 43
MarcoS said:
Good day,
I am having a similar problem/
after star aligning the registered frames, I did the star alignment based on those frames with reference image one of the registered ones.
in the resulting frames a big black area is seen beside the C-2020F3 comet.
attach few frames to show the problem.
any hints?
I ave try all the setting in the star align that I can think about with no joy....

View attachment 8821View attachment 8822
View attachment 8824
View attachment 8825
Click to expand...
I'm having the same issue. I guess the next thing for me is to try to deal with the gradient caused by the darkening background of sunset (and possible thin clouds).
 
Bump. I am having a similar issue to the above posters from last year. The subtract operand is not subtracting the comet from the calibrated images in the slightest. I know it's doing *something* because the dark band at the bottom of the images is getting slightly worse than the source images, but that's about it. I have no gradients present in my images, so the LinearFit explanation a few posts above would not seem to apply. I tried this first with a large subset of images, and when nothing subtracted in that set I tried it with only two consecutive images to reduce any potential luminance or gradient differences in the image. Both returned the same result with no subtraction of the comet. I am using the comet process described by both Warren Keller and RBA as my procedure.

Does anyone know why this would be occurring? I've been at this all afternoon looking through every tutorial possible and nothing has helped.

Pictured below, top left is the image that should have the comet subtracted, top right is the source image (there are virtually no differences between the two besides the bottom border), and the bottom image is the integrated comet master that is being subtracted.
Cometsubtract.jpg
 
Could you upload the xisf for the "comet only" and "comet + stars" images (the two images being subtracted) and post a link.
 
What I came to realise is that even if the subtraction doesn't seem to work, once you integrate the star aligned frames (comet subtracted), there will be very little or no comet remaining. This was done at Alt 30 but from a dark sky.

https://www.astrobin.com/e2p6tc/
LeonardCN.jpg
 
Did anyone have any luck with these? I'm at my wit's end. I tried linear fitting the integrated comet master prior to the subtraction and that didn't work either. There just doesn't seem to be any subtraction taking place unless I'm royally screwing something up somewhere.
 
You must log in or register to reply here.
Back
Top