Good performance of ImageIntegration

[Jordi Gallego]

Hi all,

My last image, a widefield of NGC2170 region in Monoceros, has been a bit difficult to process.
When I saw the first couple of images downloaded from the camera I thought that I had a problem in the STL: both images had a group of parallel vertical traces in the same place of the frame (over NGC2170). A more close inspection showed me that the vertical lines changed from frame to frame, so no problem on the CCD, they were really satellite traces. Astonish I saw that image after image (I collected 14x20 minutes for each R,G and B channel over November to February) all images had the traces (and all of the over the more interesting object on the field!!!). All frames had a minimum of 3 traces and a maximum of 7 on a very narrow path. You can see some of the frames (not at all the worst) here:



This a first try of a pure average combination on the green channel, showing the result.



However, after several trial and error tests ImageIntegration did a very nice job and achieved to eliminate all of them. This is, I think, a remarkable proof of the good implementation and function of the rejection algorithym (winsorized sigma clipping on this case) as it must be noted that in total I had 14 images x 3 channel x 5 average number of traces = 210 traces in a very narrow space!
Obviously I had to push the limits of the algorithym  so I also get an integrated image with a noise level higher than the one I expected with 14 exposures. Maybe because of this I am not especially happy of my processing results, but in any case this is the final result I got (and with no traces!) 8)


And this is a link to a higher resolution version:
http://jordigallego.fotografiaastronomica.com/album/Files_dark_sky_SBIG/NGC2170_widefield.jpg


Regards
Jordi
www.astrophoto.es

[Philip de Louraille]

Impressive! (both the algorithm and your post-processing!)

[martin_magnan]

Very good result Jordi. Bravo !!!

Can you elaborate on the sigma values you used for the integration please. On my last image i had only one satellite trace and it still shows in the integration using winsorised sigma.

Thanks

Martin

[vicent_peris]

Hi Jordi, beatiful image.

This is going to change very soon. We are implementing a new parameter in ImageIntegration that will raise the performance of the rejection algorithms quite a bit. This image will be a good test for the tool, and I think you will get a higher signal to noise ratio.


Best regards,
Vicent.

[Jordi Gallego]

Hi

Thank you all for your comments

Can you elaborate on the sigma values you used for the integration please. On my last image i had only one satellite trace and it still shows in the integration using winsorised sigma.

Martin the final values were Sigma low= 4.0 and Sigma high = 1.5, but each image is different and you should try to find the best figures on each case. Maybe you will find useful to take a look to this presentation on ImageIntegration,

http://pixinsight.com/forum/index.php?topic=2608.0

on pages 24 and 36 you will find a simple but systematic procedure to set the limit values.

This is going to change very soon. We are implementing a new parameter in ImageIntegration that will raise the performance of the rejection algorithms quite a bit. smile This image will be a good test for the tool, and I think you will get a higher signal to noise ratio.

This is very good news Vicent!!

Regards
Jordi

[h0ughy]

Jordi, what a fantastic result.  Thanks for the explaination fo the processing required to obtain this.  be absolutely fantastic if you could do it again by video recording it to show step by step how you obtained this result! 

[martin_magnan]

Martin the final values were Sigma low= 4.0 and Sigma high = 1.5, but each image is different and you should try to find the best figures on each case. Maybe you will find useful to take a look to this presentation on ImageIntegration,

http://pixinsight.com/forum/index.php?topic=2608.0

on pages 24 and 36 you will find a simple but systematic procedure to set the limit values.

Thanks for the link. I saw the presentation it when it was in spanish only. But now that there's an english translation, it has become a priceless document.

Thank you very much

Martin