Image Alignment tips for obstructions in frame?

[bvalente]

Hi all,

Does anyone have tips for successful Image alignment when there is a bit of land that is moving through the frame? it's a small part of the frame, but changing and present in all frames.

So far I've had good success, but the process seems to fail when it comes to terrestrial moving through the frame. I've attached an example.

If it should handle this, that's good to know too

Thanks for any suggestions

Brian

[Niall Saunders]

Hi Brian,

This is one of those situations where you, as the image composer, have to make a decision, or even a series of decisions.

There is no 'statistically justifiable' method by which you could incorporate the terrestrial data into a final image that has been based on the data from multiple aligned and stacked images. Aprt from anything else, ImageIntegration should be used to reject the terrestrial portions of the image simply because the pixels that form the image have to be considered to be 'outlier' pixels that would not conform to the acceptable image limits that II would impose during the integration phase.

So - do you then aim for the best 'astronomical' data set that you can, and avoid (or at least minimise the amount of) any terrestrial data in your data set? If so, you end up with an image that is not 'tethered' ot the ground - it has no visual cues or references to the ground and, as such, might be considered 'less appealing' for that (despite it being, scientifically, 'more accurate')

Or - do you 'add-in' the ground reference after the creation of your astronomical masterpiece? There must be many such images adorning the internet (usually with no concession given as to how they were acuired or processed).

Or - do you abandon the concept of multiple image acquisition (thus losing all the benefits in SNR that this approach brings with it), and just take a single exposure of both sky and ground? If you adopt this process,  you still have the not-insignificant challenge of processing the two wholly different image zones. Yes, you might be able to take advantage of HDRC (High Dynamic Range Compression) tools, but you are probably still going to have to invoke some form of mask that allows you to imrove each zone using entirely different image-processing methods.

I have never tried this in PixInsoght myself, and may yet find myself strapped to a ducking-stool in the local witch-ducking pond for saying so, but I might not look at PixInsight to achieve this in its own 

However - maybe others have had success where I might fail.

[bvalente]

Thanks Niall

I appreciate the input. Just to be clear i don't want the terrestrial aspects in my image. the position of the milkyway core was just inconveniently next to a mountain top, so it peaked into the frame.

I have succeeded by using photoshop to align and integrate them using a mask to mask out the terrestrial portions for alignment, but i am not able to take advantage of calibration frames in that regard

I am currently trying pixinsight 'region of interest' in star alignment to see if I can get that approach working


Brian

[Niall Saunders]

Well Brian,

Keep persevering - and let us see your end result!

[bvalente]

Success! more or less

I separated the image preprocessing (checking only "calibrate") and handled each other step individually

then used starmatch on the calibrated (and debayered) files to align them

then used imageIntegration for stacking and rejection.

One thing PI did much better than photoshop was rejection. there were some tree artifacts that PI removed that photoshop could not. Also PI removed a lot more of the mountaintop "blur" so even though I have a big black hole on the lower left, it's much smaller than otherwise, and the remaining sky around it is much more usable

This is just after histogram stretching so I haven't done any real improvements beyond integration and color calibration, but it's a solid starting point