What method does Affinity Photo use to align images + how to improve alignment for planetary sequences ?

[Didge]

Hi,
I have a question about the method used by Affinity Photo (AP) to align images.

I'm using AP (on Mac) to align 698 images of the moon taken during the recent lunar eclipse.
The images are captured as the moon moves across the sky, so there is only a small change between each image,
but a significant difference between the first image and the last one.

See the attached images ...
(a) a composite of images $001 - #698 unaligned - this shows the total spread of all the original images.
(b) a composite of images #001 - #403 unaligned - this shows the total spread of all the first 403 images.
(c) a composite of images #001 - #403 aligned - this is the first 403 images automatically aligned by AP.

As you can see, AP does an OK (not brilliant) job of aligning the first 403 images.*
But then when I add image #404, AP seems unable to align it with the existing stack of images #001 - #403.

See image (d) - a composite of (c) above with image #404 added on.

After that, AP 'loses track' of all the subsequent images - see image (e), a composite of images #001 - #698 as aligned by AP.

AP seems to generate 3 'clumps' of images that are 'sort of' aligned with each other (but not with the stack 000-403).
After that the images do not appear to have been translated at all from their original positions.

As a test, I selected images #350-#450 - they all align OK - so there's nothing 'special' about image #403.

I suspect that the moon's positions in images #404 - #698 are just too far away from its position in image #001 for AP to align, because the bright edge of the moon in image #403 has moved into the 'shadow' area of the moon in image #001.

So - I have 2 questions :-

1. am I correct in guessing that AP uses the lowest image in the stack as a reference,
    and attempts to align each other image with that as AP progresses up the stack ?

2.  *what can I do to achieve better automatic alignment than I'm getting in image (c) ?

I suspect that AP is getting 'fooled' by the increasing range of shadow in the top left corner of the moon - AP assumes that all the images are 'the same'.

The moon has a very bright and well-defined edge on the lower right-hand side - is there a way to force AP to use that region as an alignment reference, rather than the whole moon ?

Many thanks,

Gary

Image (a) - a composite of images $001 - #698 unaligned - this shows the total spread of all the original images.

Image (b) - composite of images #001 - #403 unaligned - this shows the total spread of all the first 403 images.

Image (c) - a composite of images #001 - #403 aligned - this is the first 403 images automatically aligned by AP.

image (d) - a composite of (c) above with image #404 added on.

image (e) - a composite of images #001 - #698 as aligned by AP.

[Didge]

OK - Ive done a bit more experimenting ...

I took the stack of unaligned images you can see in (a) above and cropped them to show just the bright edges of the moon.

See image (f) attached below.

Then I did a live stack (which took several hours) to see whether AP would do a better job of aligning the images with a smaller target and less black in the background.

It did improve things slightly ...
I got an OK aligned stack for images 001 - 516 (an improvement of 113 images) - see image (g) attached.
The alignment is pretty good, but not pixel-perfect.

But - when I added layer 517, AP was unable to align that frame or anything higher in the stack.
See image (h) attached, which includes the whole stack.

Any suggestions on how to get better alignment with respect to the bright edge would be very welcome.

Regards,

Gary

=====

Image (f) - stack of 698 unaligned images

Image (g) - images 001 - 516 as aligned by AP - an improvement of 113 images

Image (h) - The stack from (g) above with images 517 - 698 included - AP was unable to align these.

[Didge]

OK - Ive done one last experiment ...

I took the first 514 images of the stack from (g) above, and did a VERY tight crop to include only the bright area of the moon - see attached image (i).

Then I cropped and rasterised each image, and re-exported them as TIFFS.

Then I created a new stack with AP set to auto-align.

I did expect AP to do a much better alignment this time, because the images being aligned are consistently illuminated ...
hoping AP wasn't being 'fooled' by the dark area of the moon becoming progressively larger with each shot.

Unfortunately, the result was pretty much identical - as you can see in attached image (j).

The aligned stack is identical to the unaligned one.

 I'm not sure what this tells us about the alignment method ...

Can anyone suggest a way to get better alignment based on using such areas of high contract ?

Gary

=====

Image (i) - Very tight crop of images 001 - 514 > each exported as TIFF.

Image (j) - Images reimported as TIFFs and auto-aligned - result is the same as for image (i) above.

=====