Flip cutout images by snbianco · Pull Request #180 · spacetelescope/astrocut

snbianco · 2026-04-21T21:12:07Z

When creating PIL.Image objects or writing cutouts to image formats, vertically flip the image to match the intended orientation of astronomical data.

Added a flip_orientation parameter to get_image_cutouts and write_as_img to allow users to opt in or out. The default is True, but I wanted to keep the legacy option of the "unflipped" version.

scfleming · 2026-04-21T21:15:11Z

If this is updating the orientation of the cutouts, how was this missed the first time? Was our comparison between old and new not sufficiently detailed to include orientation checks, along with pixel-level checks I am fairly certain were done? Or was the testing focused on the FITS output and not sufficient coverage of manual testing on the JPG output?

rlwastro · 2026-04-21T21:17:27Z

@scfleming The orientation problem affects only JPEG (or other graphic formats). My guess is that the existing checks did not check the orientation on those images. Checking that the FITS cutouts are correct at the pixel level would not pick up this problem.

snbianco · 2026-04-21T21:17:50Z

If this is updating the orientation of the cutouts, how was this missed the first time? Was our comparison between old and new not sufficiently detailed to include orientation checks, along with pixel-level checks I am fairly certain were done? Or was the testing focused on the FITS output and not sufficient coverage of manual testing on the JPG output?

Historically, the orientation of image outputs has never been flipped. The new and old version output the same thing, but they were both incorrect.

codecov · 2026-04-21T21:18:26Z

Codecov Report

❌ Patch coverage is 98.30508% with 1 line in your changes missing coverage. Please review.
✅ Project coverage is 95.96%. Comparing base (d993436) to head (58f31f2).

Files with missing lines	Patch %	Lines
astrocut/image_cutout.py	98.30%	1 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #180      +/-   ##
==========================================
+ Coverage   95.92%   95.96%   +0.04%     
==========================================
  Files          20       20              
  Lines        2232     2279      +47     
==========================================
+ Hits         2141     2187      +46     
- Misses         91       92       +1

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

scfleming · 2026-04-21T21:22:25Z

OK roger that, a good point to take stock of what we test as we develop, that's all. Glad we are getting this fixed with good tests, etc now.

snbianco · 2026-04-21T21:25:20Z

I wasn't fully aware of the convention for flipping images, so thank you @rlwastro for letting me know! I added some assertions to the tests that check that the image was flipped.

snbianco · 2026-04-22T13:45:15Z

+
+
+@pytest.mark.parametrize("colorize", [True, False])
+def test_fits_cutout_img_flip(test_images, center_coord, cutout_size, colorize):


Most of the changes in this file are stylistic. This is the new test!

snbianco · 2026-04-22T13:46:22Z

-from pathlib import Path
-from typing import List, Optional, Union, Tuple
 import warnings
+from abc import ABC, abstractmethod


There are a lot of style changes in this file as well. The relevant changes have to do with the flip_orientation parameter.

havok2063 · 2026-04-23T13:49:16Z

Do we embed any metadata into the png image cutouts? Like coordinates or wcs or anything like that? If not, we may want to consider that. It may help with verification of correct orientation in the future, or allow users to overlay onto other things.

snbianco · 2026-04-23T15:43:44Z

TIL you can add metadata to images! I love this idea. I added a few keywords that were pretty trivial, but I think something like a wcs (or gwcs in the case of ASDF) would require a little more work and may be outside the scope of this PR. This is something I will add to the backlog though.

havok2063 · 2026-04-23T16:00:33Z

+        meta.update(
+            {
+                "center_ra_deg": self._coordinates.ra.deg.item(),
+                "center_dec_deg": self._coordinates.dec.deg.item(),
+                "origin": "STScI/MAST",
+                "version": __version__,
+            }


should we also encode the pixel scale used in the cutout? Or cutout size or some info about the boundaries?

Added! You can also access the dimensions through the width and height attributes of the Image object, but I figure we can make them more obvious.

havok2063 · 2026-04-23T16:04:27Z

I added a few keywords that were pretty trivial, but I think something like a wcs (or gwcs in the case of ASDF) would require a little more work and may be outside the scope of this PR. This is something I will add to the backlog though.

Yeah, I embedded a standard FITS WCS for some png image cutouts with reasonable success. It was accurate enough to reasonably recover and overlay the image onto other FITS. The one caveat is the conversion to/from can be tricky since you have to convert every field to a string. gwcs might be trickier, but may also not be needed for a simple image cutout.

scfleming · 2026-04-23T16:05:12Z

TIL you can add metadata to images! I love this idea. I added a few keywords that were pretty trivial, but I think something like a wcs (or gwcs in the case of ASDF) would require a little more work and may be outside the scope of this PR. This is something I will add to the backlog though.

I may be incorrect about this, but I thought the older PanSTARRS cutout service provided JPEG output with metadata/WCS embedded in them? I am fairly certain something at MAST had/has this capability, and Rick might know about it....just FYI, I agree adding WCS is a great idea and could be very useful especially with some of the stuff Cobalt team is doing (they might also have capability of exporting JPG with WCS info embedded for overlaying on Aladin), but as a separate PR.

rlwastro · 2026-04-24T15:13:03Z

@scfleming Scott is correct, the current fitscut service does add the WCS to JPEG images in the JPEG comments section. This was originally added because it was supported by Aladin, which can display a JPEG with an embedded WCS. It turns out to be pretty easy to access from within Python using PIL.
I do think this is worth doing, even if it is not essential for this particular change. The WCS consists of just strings with the card images (without trailing blanks) for the FITS WCS header keywords.
Probably C code for this is not very helpful, but here is a link to the code in fitscut that adds the WCS header.

scfleming · 2026-04-24T15:39:20Z

when we add the WCS in the output, you can check with Brett on how they are handling the aladin overlay with Roman gwcs, e.g., I think they are using an approximation of gwcs to handle what FITS supports, which is probably sufficient for use within Aladin especially over a small area, but could to check in with them what they've learned since they are also working on this for exporting JPG from Jdaviz as overlays on Aladin. We'd want astrocut JPG output to have same precision and implementation as what's being done for Jdaviz I think.

snbianco marked this pull request as ready for review April 21, 2026 22:01

snbianco commented Apr 22, 2026

View reviewed changes

snbianco requested a review from havok2063 April 22, 2026 13:53

havok2063 reviewed Apr 23, 2026

View reviewed changes

snbianco added 4 commits April 30, 2026 14:38

Flip cutout images, test cases, docs update

b539b8c

changelog, add to write_as_img

2f4e04c

Image metadata

03d8cee

Add cutout size and pixel scale

58f31f2

snbianco force-pushed the flip-jpeg branch from 7eb6bb2 to 58f31f2 Compare April 30, 2026 18:40



		@pytest.mark.parametrize("colorize", [True, False])
		def test_fits_cutout_img_flip(test_images, center_coord, cutout_size, colorize):

Conversation

snbianco commented Apr 21, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

scfleming commented Apr 21, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

rlwastro commented Apr 21, 2026

Uh oh!

snbianco commented Apr 21, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

codecov Bot commented Apr 21, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Codecov Report

Uh oh!

scfleming commented Apr 21, 2026

Uh oh!

snbianco commented Apr 21, 2026

Uh oh!

snbianco Apr 22, 2026

Choose a reason for hiding this comment

Uh oh!

snbianco Apr 22, 2026

Choose a reason for hiding this comment

Uh oh!

havok2063 commented Apr 23, 2026

Uh oh!

snbianco commented Apr 23, 2026

Uh oh!

havok2063 Apr 23, 2026

Choose a reason for hiding this comment

Uh oh!

snbianco Apr 23, 2026

Choose a reason for hiding this comment

Uh oh!

havok2063 commented Apr 23, 2026

Uh oh!

scfleming commented Apr 23, 2026

Uh oh!

rlwastro commented Apr 24, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

scfleming commented Apr 24, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants

snbianco commented Apr 21, 2026 •

edited

Loading

scfleming commented Apr 21, 2026 •

edited

Loading

snbianco commented Apr 21, 2026 •

edited

Loading

codecov Bot commented Apr 21, 2026 •

edited

Loading

rlwastro commented Apr 24, 2026 •

edited

Loading