not a bug Strange warping in corners with fast-integration

[sbock1]

Hi,
I just used WBPP and fast integration for processing pictures I took of IC410 . I was shocked to see a strange "warping" effect (most visibly at bottom left, see below) in the final image. I first thought something with the image train was wrong, but I couldn't think of anything that caused this behavior. Then I ran WBPP again, but now without the "fast integration" box checked (otherwise identical settings). That resulted in a good stacking without warps in the corners.
I am wondering if this is caused by some bug in the algorithm or if it is caused by the nature of the images which hat bloated stars due to bad seeing conditions.
I used a Canon DSLR (146 frames, 1minute each) and PI 1.8.9-2 build 1596 on an Windows 10 machine.
I would love any hint how to avoid this because it forces me to not use "fast-integration" and, therefore, spend a lot more time
I can provide the raw images or try out things if it helps to determine the problem
Best regards
Stephan

Image stacked with fast-integration (warped corner bottom left, but also top right):

Picture with "conventional" registration+integration

Closeup 1:1 of bottom left corner of warped picture.

 

[robyx]

This may occur when the location of the reference stars is not well distributed in the image, as well as if you don't have enough reference stars to guarantee a good global alignment over the whole image.

You can see which are the reference stars by opening your reference frame, setting the "Reference stars" Working Mode and dragging the square triangle from the process to the image, this will show the selected reference stars:

One suggestion is to reduce bth the search box size and the iteration expansion to decrease the search region and so increase the number of reference stars:

Let me know if this helps.

Robyx

 

[robyx]

Moreover, if you want to inspect how the registration worked, you can export the registered images by checking the "Generate registered images" checkbox and setting a valid output directory:

Once done you can blink the images and see how the fast registration did work, retune your parameters, and run again to see if the registration improved.

 

[sbock1]

Hi Robyx,
thanks for the hint. I looked at the reference stars and they are not all the way in the corners indeed. However, I have 60 stars, even with standard settings. As a lay person, I would say that the distribution doesn't look too horrible (see below). If I increase the amount of Reference Stars to 128, the result does improve, but still shows artifacts in the corners.

So, that begs the question, how to move the reference stars into the corners and how many are enough?
I was able to recreate the effect with only 4 pictures, so I can test quickly.
Should I find a suitable setting, is there a way to use that in WBPP?
Thanks
Stephan

 

[sbock1]

Ok, so I found a way to exclude most larger stars by reducing "Star Detection" "Upper Limit". Then only the small stars are detected, but those are distributed more evenly. Also, setting the "Reference Stars" to 128 and reducing the search box as you suggested gave me a decent distribution.

I tested it and the warping is gone. However, even with "Max median tolerance" increased to 3.0, I still had a success rate of only 78%. Probably due to the faintness of the reference stars.
Thanks for the pointers, I assume this is no "bug" then, but simply a result of the star distibution in the picture.
Nevertheless, is there a way to use WBPP with these settings?
Stephan

 

[robyx]

Ok, so I found a way to exclude most larger stars by reducing "Star Detection" "Upper Limit". Then only the small stars are detected, but those are distributed more evenly. Also, setting the "Reference Stars" to 128 and reducing the search box as you suggested gave me a decent distribution.View attachment 21749
I tested it and the warping is gone. However, even with "Max median tolerance" increased to 3.0, I still had a success rate of only 78%. Probably due to the faintness of the reference stars.
Thanks for the pointers, I assume this is no "bug" then, but simply a result of the star distibution in the picture.
Nevertheless, is there a way to use WBPP with these settings?
Stephan

very well done!