Comet NEOWISE - A request for help

[Tim Ellison]

Hi Everybody,

I wondered if I might get away with coming back to the forum with a second "begginerish" question. I hope that's OK!

I have been having a go at processing an image of Comet NEOWISE. I don't think I am the only person who has been doing that! So I'm hoping that somebody might have got on rather better with it than I have, and might have a solution to a problem that I find myself stuck with.

Here's the story. I have 61 decent subframes, taken using a DSLR and lens and a Star Adventurer tracking mount. I also have dark, bias and flat frames to go with them. Having successfully calibrated the subframes, I have then used the "CometAlignment" process. That seems to work very nicely and does a great job. But I have run into a problem when I try to integrate the "comet aligned" set of frames. I end up with ghostly star trails in the result. I have spent a lot of time tinkering with the "ImageIntegration" process in the hope of getting rid of them. I've tried several different rejection methods and played around quite extensively with the settings in each. But whatever I try the star trails are still there. I have found ways of making them fainter by adjusting the integration settings, but they don't ever completely go away.

Two images are shown below, for illustrative purposes (reduced quality copies). The first is one of my original raw subframes, with just a screen transfer function applied. The second is the best "finished" image I have been able to produce so far - note the ghostly star trails. And the trees in the corner (a separate issue!).

I am wondering what I might be able to do. At the moment I can think of four possibilities (apart from giving up!) :-

1. I've tried using Large-Scale Pixel Rejection, but am not very sure what I am doing with it, and haven't been able to make it do anything useful. I wonder if there might be a write-up anywhere which explains its principles of operation, and might allow me to judge whether it would be appropriate to use, and how to work out what settings to apply.

2. I'm wondering about Local Normalization. It seems it can be helpful where strong background gradients are present, which they are in my data. But I've also seen warnings against using it. Is anybody able to offer any advice or guidance? Might this allow the pixel rejection aspects of image integration to work better?

3. I'm wondering whether I should try to remove the stars from the comet-aligned frames before I integrate them. Does anybody have an opinion on this? Is it a silly idea? If not, what might be a good way to go about it?

4. I have been performing thought experiments involving PixelMath. I wonder if there might be a "clever" way of combining the star-aligned and comet-aligned images which would get rid of the star trails. So far I haven't managed to come up with anything that looks likely to work.

Does anybody have any ideas or advice?

Many thanks in advance in case the answer is "yes".

Regards,

Tim

 

[pfile]

the star residuals are not necessarily uncommon - stars have a gaussian shape with a faint area surrounding them, and where they overlap the rejection algorithm thinks the data is real.

since you have 61 frames, one thing to try would be to break up the comet-registered images into groups, with spacing between the subs large enough to make sure there is little to no overlap in the stars. so for instance if you find that image 1 and image 5 have stars far enough apart, then you make 12 stacks with images 5 apart from one another, then integrate all the results together.

rob

 

[ngc1535]

Tim,

This is something I demonstrate in my tutorials. I go about this stuff in weird ways.
(I think that gets me into "trouble" since it isn't common solution, although the problem certainly is common.)

You know something about the star residuals right? They are coming from the almost rejected stars in the comet aligned images. So.. if you add (not reject!) all of the comet aligned images you will have an image with the exact trails that correspond to residuals you see.
Then on this image use MLT at a small scale size to create a structure map of just the star trails (the comet is larger than the stars so it will not show up). Probably just the first four layers. Now you have the beginnings of a mask of your residuals. You can binarize this MLT'ed result and have a ready to go mask. Apply this mask to your comet aligned /integrated image (probably with a little dilation and convolution) and further mitigate the residuals with a combination of MMT and noise generation. Then viola! you have the result you were looking for.
At least..that is what I did with my image.
-Adam Block

 

[Tim Ellison]

Hi Rob and Adam,

Those are interesting ideas and I can see the sense in both. Thank you for taking the time to explain them. I'll have a try and see how far they take me.

Thanks again.

Tim

 

[Tim Ellison]

Sorry I have taken quite a while to come back on this. Life got in the way for a while and I have only recently been able to get back to image processing. But I have now spent a decent chunk of time looking again at the comet image. I tried the idea of splitting the frames up into batches for integration. It sounded like a great idea, but sadly I haven't been able to make it work. Even splitting into 15 batches of 4 frames each didn't get rid of the trails. I had a close look at the raw frames and can see that the stars all have a (very faint) glowing aura on one side of them. I suspect it is an imperfection arising in my camera lens. Anyway, it looks like that is the root cause of the problem - the starlight is, as you suggested Rob, spread over quite broad area, which defeats any attempts at rejection during integration. At least I now think I understand the problem better. Adam - I started having a go with your suggestion, but got out of my depth fairly quickly with it. I think I'm still a bit too much of a beginner! I might need to get hold of your tutorials and do the learning properly. Thank you both again for your help. I'm sorry I can't (yet) show you a finished image! Regards, Tim

 

[txastro]

Tim,

This is something I demonstrate in my tutorials. I go about this stuff in weird ways.
(I think that gets me into "trouble" since it isn't common solution, although the problem certainly is common.)

You know something about the star residuals right? They are coming from the almost rejected stars in the comet aligned images. So.. if you add (not reject!) all of the comet aligned images you will have an image with the exact trails that correspond to residuals you see.
Then on this image use MLT at a small scale size to create a structure map of just the star trails (the comet is larger than the stars so it will not show up). Probably just the first four layers. Now you have the beginnings of a mask of your residuals. You can binarize this MLT'ed result and have a ready to go mask. Apply this mask to your comet aligned /integrated image (probably with a little dilation and convolution) and further mitigate the residuals with a combination of MMT and noise generation. Then viola! you have the result you were looking for.
At least..that is what I did with my image.
-Adam Block

Adam,
Is this the same as your star de-emphasis procedure?

Thanks,
Xavier

 

[ngc1535]

Adam,
Is this the same as your star de-emphasis procedure?

Thanks,
Xavier

No... this is something different. This is really just a clever way to generate a mask by using the "star trails" which are the precise positions of the residuals of stars that are not completely rejected with image integration. Then with MMT and noise generation you clobber the star residuals...
-adam

 

[txastro]

No... this is something different. This is really just a clever way to generate a mask by using the "star trails" which are the precise positions of the residuals of stars that are not completely rejected with image integration. Then with MMT and noise generation you clobber the star residuals...
-adam

Thanks! The procedure descriptions had similarities, and I was curious.

Xavier

 

[ngc1535]

Thanks! The procedure descriptions had similarities, and I was curious.

Xavier

Well..yes, you are correct that the use of MMT and noise generation is an element of both techniques and *that* really points to how powerful the idea is.
Good job!
-adam