Upside down images?

Started by nebulachadnezzer, February 12, 2021, 06:49:31 PM

Previous topic - Next topic

nebulachadnezzer

My images are all rendering in Observatory flipped. None of my other viewers or editors are doing this including PixInsight.

This seems to be occurring with all of my images, captured from various ZWO cameras by both ASIAir Pro and Kstars/Ekos.

The matching, tagging, and overlays are all messed up as well. The only fix seems to be to manually flip them and then force a blind match against astrometry.net even though I have downloaded the recommended catalogs and pointed Observatory to them.

Here's an example. The first link is the image Observatory uploaded to astrometry.net using my API key. The second is the one I manually uploaded of the same file.

Upside-down, courtesy of Obseratory:
http://nova.astrometry.net/user_images/4359122#annotated

Correct orientation, courtesy of me:
http://nova.astrometry.net/user_images/4359359#annotated

Also here is what two example images in Observatory look like. I've told it to flip these manually. The first one is the automatically matched version showing the annotations are all in the wrong location and wrong scale. The second is the result after a blind match with astrometry.net. Notice the scale in the lower left and the orientation in the lower right.

Original matching by Observatory, wrong: https://imgur.com/a/JnzYUD7

Updated matching via astrometry.net, correct: https://imgur.com/a/ff97yMi

Any ideas? This software appears to be great EXCEPT for this major issue making it essentially unusable.


Sander Berents

The FITS standard doesn't specify if the origin is at the bottom-left (Y going up), as is used in math and physics, or at the top-left (Y going down), as is often used for computer graphics. There is no right or wrong here, and early on in the development of Observatory it was decided to use the bottom-left. That caused a problem though when support for XISF (PixInsight native format) was added, as this format does explicitly specify the top-left as origin. So to at least be consistent when displaying an image in FITS or XISF format, the image is indeed flipped for XISF. I should have just changed the origin back then already, as this has been brought up several times. I am indeed considering switching the origin to the top-left, as it is more consistent with other software, and consistent with XISF, but I didn't make up my mind yet for the least impactful approach.

As for your plate solving issue, I doubt it is related to the image orientation. The plate solver is known to have issues with wide field images, and this image covers approximately 10 square degrees.
Sander Berents
Code Obsession, LLC
https://codeobsession.com

nebulachadnezzer

Hi. To be clear, I am working with FITS files here, not XISF files. I only mentioned PixInsight as a somewhat authoritative source on the matter. Literally every other piece of astrophotography software I've encountered on Mac, Windows, and Linux defaults to top-left as the origin for FITS data. Only Observatory is rendering them with a bottom-left origin.

You mentioned you're unsure of the least impactful solution, and I suppose that could be because it would impact existing users who are happy with bottom-left as the default.

However, couldn't you simply offer a switch in preferences to default to top-left instead of bottom-left?

Regarding plate solving, is there any way to get Observatory to suppress that upon importing the images and only perform plate-solving later? Or perhaps just get it to accept the coordinates in the FITS header? I know these can often be wrong, but if I am taking care to ensure they are right I hate to see Observatory ignoring them, taking forever to import my images as it solves each one, and then get it wrong.

Basically your app is EXACTLY what I want to manage my library but these issues are fatal to my intended use.

Sander Berents

For FITS, the standard doesn't specify. In PixInsight it is an option, but it defaults indeed to the top-left.

A toggle in the preferences won't work because it cannot control the behavior of the Quick Look plugin. So if we switch to top-left, it is not going to be a toggle, but a change of behavior for the application and the Quick Look plugin. It's not a minor change, and impacts all existing users, which is why I haven't done it yet.

Observatory doesn't run the plate solver when importing. When it finds WCS header information in FITS and XISF files when importing it will tag the images, adds the badge in the browser and displays the overlays, but it will not run the plate solver. Some acquisition software adds incorrect or incomplete information to the FITS headers, and this will result in incorrect tags and overlays. In that case, select the images and choose Image > Unmatch in the menu. It will keep the coordinates and image scale information, so you have a starting point for a clean plate solve.
Sander Berents
Code Obsession, LLC
https://codeobsession.com

nebulachadnezzer

I did try converting these files to XISF by just doing a load and save in PixInsight and they still seem to render upside-down. I was hoping for a bit of a quick solution there but no such luck. This also broke the pixel scale as far as Observatory sees it pretty badly.