"Normal" Images Become Warped on Integration

[timtrice]

Attached is the CSV output from Subframe Selector. I'm also including a link to the integrated FITS image.

The problem seems to be somewhere in registration. All of my images, RAW and debayered FITS appear normal. However, when I go to register them they are not aligning correctly. I've ran registration several different times with different parameters but I can't quite get this to straighten up. In some instances the registered images becomes very skewed or tilted.

Is this something that has been seen before? What should I look at to isolate the problem and correct it?

All images were taken on the same night back to back. Other than one image which was a capture at 800 ISO, 240 seconds, all are 300 seconds.

Integrated image (no rejection): https://s3.amazonaws.com/timtrice-pictures/astrophotography/wth.xisf

I'm also attaching the registration output.

Thank you

Tim Trice

[andrewluck]

Hi Tim

Are you sure about your choice of reference image? It appears to have less than half the number of stars as the others and FWHM looks suspect as well.

Regards
Andrew

[timtrice]

Andrew,

It's done it on a number of different reference images. Attached is the registration output with the reference image of 22h44m. Integration just produces a slightly different final image but it's still warped significantly.

[Andres.Pozo]

Hi Tim,

have you tried to activate the hot pixel removal and/or the noise reduction functions of StarAlignment? The integrated image from your first post has a lot of uncorrected hot pixels that could confuse the alignment algorithm.

It could also be useful to run CosmeticCorrection before aligning the images.

Could you upload the reference image and another one which aligns badly?

Andrés.

[timtrice]

Andres,

I'm not able to reproduce the warped images for some reason now though the settings hadn't changed. Not sure what I'm missing now.

I went ahead and gzipped all 15 debayered (but not registered) images here: https://s3.amazonaws.com/timtrice-pictures/astrophotography/M3+-+2016-07-07+-+debayered.tar.gz

This morning I've been fiddling with the detection scales (increase), hot pixel removal (from 2 to 3) and log from -1 to 0 now down to -0.5. None of the results have been any different.

Is there a possibility something went wrong in my debayering? I went back and looked at those settings but saw very little to adjust.

What's confusing me most of all is I have this object imaged from another session (different framing) and I had no issues getting a good integration out of that, generally with the default setting. When I first ran this new framing through integration I did winsorized default and that's when I noticed the issues. And I haven't been able to isolate the problem since...

Tim

[timtrice]

I looked through the star alignment output and isolated images that had peak delta errors above 1px. This leaves me with 11 images. When I run those through the II process the image comes out significantly better. I haven't measured eccentricity but there is no trailing or radial warping around M3.

So something I'm doing in the registration process (or in my four "bad" images) is causing the peak delta error on one or both axis to be greater than 1px. What setting or model can I look at to improve this?

Image is attached.

[Andres.Pozo]

I have downloaded your data (I only needed two images) and as I said your problem is the hot pixels. You REALLY want to remove them before debayering and aligning.

Despite this, I have been able to align your images. I used the default parameters with these changes:
Noise scales: 1
Hot pixel removal: 2
Noise reduction: 1
Log(sensitivity): -0.4

This last parameter is very important: increasing its value it reduces the sensitivity of the star detector.

When in doubt, it is usually very useful to set the working mode "Detected stars" in order to check if the algorithm is detecting true stars or hot pixels.

The attached image is the result of executing AberrationInspector on the integrated image.

Andrés

[timtrice]

Andres,

Thank you for your help with this. I attempted several different cosmetic correction runs; none of which were producing any better results. This morning I went back and captured some darks and made a synthetic flat then started the process over again. I was finally able to get a good stacked image.

Thank you for pointing me in the right direction.

Tim