Details
-
Type:
Story
-
Status: Done
-
Resolution: Done
-
Fix Version/s: None
-
Component/s: pipe_tasks
-
Labels:None
-
Story Points:1
-
Epic Link:
-
Sprint:Science Pipelines DM-W16-5
-
Team:Data Release Production
Description
DetectCoaddSourcesTask.scaleVariance is used to adjust the variance plane in the coadd to match the observed variance in the image plane (necessary after warping because we've lost variance into covariance). The current implementation produces the wrong scaling in cases where the image has strongly variable variance (e.g., 10 inputs contributed to half the image, but only 1 input contributed to the other half) because it calculates the variance of the image and the mean of the variance separately so that clipping can affect different pixels.
Getting this scaling very wrong can make us dig into the dirt when detecting objects, with drastic implications for the resultant catalog.
Attachments
Issue Links
- mentioned in
-
Page Loading...
Hey Lauren MacArthur, would you mind reviewing this please? It's pretty much a straightforward cherry-pick from the HSC tickets above. Code on pipe_tasks PR #26. Successful Jenkins build #7273.