Talking about satellite tracks


Hi, I just want to share my experience in removing satellite tracks from frames.
I had two subframes with strong tracks:

(already calibrated, CC, weighted, registered, drizzle and ln)

First Image Integration run applying Avereged Sigma Clipping (9 subframes) and default sigma values:
tracks are still well visible.

Best result obtained after tweaking sigma high to 1.5:
better, but still visible.

New approach, following this post: and using Generalized Extreme Studentized Deviate with 0.77 ESD outliers and 0.79 ESD significance:
better, but tracks are still there and general quality of the image is deteriorated

Last chance, using Robust Chauvenet Rejection algorithm with RCR limit 0.8, range low 0, range high .98:
Finally tracks are out, but image has a lower SNR.

Conclusion: in this case at least, Yes you can remove tracks from your image, but you have to balance that with the lower SNR level of the integration. On a stack of 9 subframes my choice is to go with RCR, and when I run integration also with subframes from other session (46 images) I simply drop the two with the tracks.

This is my very, very little wisdom pill on satellite tracks removal. I would like to know other experience and learn from that.


  • track01.jpg
    20.3 KB · Views: 64
  • Track02.jpg
    30.8 KB · Views: 64
  • int_std.jpg
    26.6 KB · Views: 66
  • int_tweak.jpg
    26.8 KB · Views: 65
  • ESD.jpg
    32.8 KB · Views: 66
  • RCR.jpg
    32.5 KB · Views: 61
My experience is quite different: I never had any problem with satellite tracks rejection using the Generalized Extreme Studentized Deviate algorithm.
The difference with your experimentation is that I always had at least 40 images to integrate. I believe a sufficient number of frames is the key to a successful rejection and I would suggest not to drop any images for your 46 images session.
Thank you Nico. When, in your experience, do you put together subframes from different nights? If I understand well your point I should unite all the subframes and run a single Image Integration session. Or is it better to prepare several drizzle integration, one per night (4 in my case), and after that run Image Integration on the 4 drizzle integration frames?
Yes I believe it's much better to do a single Integration of all your calibrated (and registered) images and I think it will optimize the rejection. The workflow would be something like this:
-You calibrate your frames by nights (with a master flat for each nights)
-You register all calibrated files to a single reference frame
-You integrate all the registered files
-You do a drizzle integration

All these operations can be (and should be) done in a single instance of WBPP
Last edited: