DM-45361: Improve maskStreaks line fitting and mask plane behavior #363
Conversation
| @@ -22,6 +22,8 @@ | |||
| import numpy as np | |||
|
|
|||
| import lsst.afw.detection as afwDetection | |||
| import lsst.afw.image as afwImage | |||
| import lsst.afw.math as afwMath # FOR TESTING BINNING | |||
There was a problem hiding this comment.
Do we still need the clarifying comment since afwMath is necessary for making the binned image?
| self.config.streakBinFactor) | ||
| binnedExposure = afwImage.ExposureF(binnedMaskedImage.getBBox()) | ||
| binnedExposure.setMaskedImage(binnedMaskedImage) | ||
| binnedExposure.setPsf(difference.psf) # exposure must have a PSF |
There was a problem hiding this comment.
So I have been informed that there is a possibility that you can reach this point in the pipeline and an exposure might not have a PSF. How does that get handled in this situation? Does this automatically fail out gracefully or do we need to add some error handling just in case.
There was a problem hiding this comment.
Thank you for thinking of this — indeed this was a past problem, so I verified that the _prepareInputs call in the run method of detectAndMeasure will raise well before we get to streak masking if the PSF is un-compute-shape-able.
| self.assertTrue(np.all(streakMaskSet[streak_check])) | ||
| # Check that the entire image was not masked STREAK | ||
| self.assertFalse(np.all(streakMaskSet)) | ||
|
|
There was a problem hiding this comment.
If error handling is needed in the main code, then you should include a null test where things go wrong so you can check that the error is properly caught.
There was a problem hiding this comment.
Not relevant since we can never land here without a valid PSF
mrawls
force-pushed
the
tickets/DM-45361
branch
from
abf95ee to
2f3b1b4
Compare
This PR adjusts how maskStreaksTask is implemented in difference imaging. It uses a mask plane binning strategy to more readily detect fainter streaks, thereby "connecting the dots" in the DETECTED mask plane before looking for linear features.