Image Integration (Comet Processing) gone awry

[XcalRocketMan]

Having purchased the wonderful book "Inside PixInsight" by Warren Keller, I had the opportunity to use the CometAlignment process of PI to isolate 20 subs of comet 45P I took a couple of evenings ago, producing a set of files of just the comet and a set of files with just the star field, planning on producing a final image of 45P against a non-trailed starfield. Following the procedure in the book to the letter I find I am at a conundrum.

Whereas the ImageIntegration of the star field produces an acceptable master, the resultant master from the ImageIntegration of the comet registered files yields a master that appears to be non-linear (and stretched quite a bit). The background is very bright with mean pixel values of 0.7 and higher whereas the master of the star field background is only 0.017, which is rather normal for my area. It seems as if the CometAlignment process (when the subs shift) produces a non-linear image. I tested this by applying CometAlignment without selecting any location points (which might actually prevent the real process from executing, and so my test may be invalid) and the result of integrating those images is fine.

The issue seems to be in the normalization process of image integration (A fellow AP'er was kind enough to figure this out for me). "Additive" and "Additive with scaling" causes the odd result; "no normalization" and "multiplicative" seem to work fine.

Sample files can be downloaded from:  https://www.dropbox.com/sh/gd3nj04lcs5higd/AADETi9OHi8OIYl9qsPSqma0a?dl=0

Mikey

[Stu]

I see your post here.  Replying so as to be able to see responses. 
Stuart

[Olivier]

Hello,
I am experiencing the EXACT same issue, with stacking the comet aligned subs.
(I was imaging exactly the same 45P, the only difference is that it was on Monday night 02/20/17)
I just figured the same thing, that is the problem apparently coming fron the normalization step.
If anyone interested, I can post the subs on dropbox.
Olivier

[EastWind]

Seeing no other input here, i will chime in as a newbee who also has a lot of frustrations with comet align.  I cant say for sure if its non linear or just additive due to the way it stacks the comet and applies pixel rejection to the rest.

It was best if i used linearfit to align the comet integration to the star integration before using pixelmath to combine the two images.

Also with the new cometalign i found GREAT success with the seldom documented subtraction options.  Use the first cometintegration as a sort of mask to creat the star integrated background.  Use clone stamping to get rid of most of star trail artifacts in the cometintegration.  Then use that to create the starintregrated background by subtracting it. THEN do the cometalign a third time and subtract the star background you just created.  This gives you a comet integration without any star artifacts.  Then linearfit and pixelmath to combine both.

The first cometintegration is used only to create the starstacked background and then never used again.

 Most of the books out there are a bit dated and describe the old school method of comet stacking which requires a great deal more manual labor to get a clean image.  Its still a great idea to go through the paces as it helps to understand the core principles.  However i can tell you that unless you have a fantastic comet signal it will be difficult to process out a lot of the background artifacts.  The subtraction method gets rid of most artifacts and those that are left are darker than the background as you are subtracting them and easier to get rid of.

I am planning to make a short video on how to use subtraction as there is not much out there yet on it.  Took me a while to decode the existing forum discussions and understand what PI was doing and thats simply because i am fairly new to PI and still learning. 

Now im waiting for 41P and the moon cycle to end so i can get out and start imaging again.