Resample validation testing

[stscijgbot-jp]

Issue JP-1531 was created on JIRA by Alicia Canipe:

resample

• Update to return resampled data in an ImageModel, instead of DrizProductModel. [#⁠4552]
• Updated documentation to include step arguments and reference file description. [#⁠4723]
• Change outlier and resample DQ bit usage. [#⁠4726] The parameter good_bits has been removed in favor of allowing all DQ flags except for DO_NOT_USE
• Updated to reject pixels with DQ flag NON_SCIENCE, in addition to DO_NOT_USE. [#⁠4851]

The updates above were made to the resample step of the level-2b and 3 pipelines. Perform validation tests for your instrument to check that the step is working for all modes and configurations, and that it is performing as expected according to the algorithm defined by the CalWG.

[stscijgbot-jp]

Comment by Misty Cracraft on JIRA:

The resample notebook stored in the jwst_validation_notebooks github repo for miri resampling tests this step insofar as finding the stars and matching them to the expected ra/dec coordinates. If those stars are found to within a threshold distance of the expected coordinates, the test passes. That notebook reads the resampled image in as an ImageModel, so that has been confirmed to work correctly. While we don't have a DQ extension for the resampled image, I looked at the flat file DQ array for pixel regions labeled 'NON_SCIENCE', and confirmed that the 4QPM areas in the imager are flagged as NON_SCIENCE, and those regions are 0 in the combined resampled image, which shows that the resample step is correctly rejecting those pixels and not trying to combine them in the output image.

So the issues listed above in the description have been tested and shown to be working as expected. The resample notebook covers further aspects of this testing, namely that the stars are at the expected ra/dec coordinates.

There are some image anomalies in the combined data that might require further analysis.

In the attached image, the lighter grey areas, the 4QPM region, have pixel values of 0, as expected. The background values around the stars are all ~ -10, which indicates that the background was subtracted twice (already filed in another Jira ticket), and the regions between the Lyot and imager portion are around -20, while the lyot bar and circles are either around -15 or -20. There are still some questions for me about how the image combination is being done in those regions. Also, there are some odd streaks coming out from the imager in the lower portions of the image, and I'm not sure what those are.

 

[stscijgbot-jp]

Comment by Kevin Volk on JIRA:

Testing of the resample step was done for individual NIRISS images in the image2 pipeline.  The resampling appeared to preserve the photometry and astrometry of the sources.  See https://jira.stsci.edu/browse/NIRISS-123 for a brief description of this test.  A separate comparison was made for the combined resampled image from the image3 resample step, which showed issues (see tickets JP-1570 and JP-1579) which probably indicate issues in the resample step (I do not know if there is a distinction between image resampling for individual images, which works, and image resampling to combine multiple images, which has issues, in terms of the internal workings of the step).  Work on this is described in https://jira.stsci.edu/browse/NIRISS-156 .

[stscijgbot-jp]

Comment by Alicia Canipe on JIRA:

Build 7.5 testing is complete. Closing the related issues in preparation for Build 7.6 testing.