Uploaded image for project: 'Data Management'
  1. Data Management
  2. DM-32355

Add function to remove Poisson contribution from sources to variance plane to meas_algorithms

    XMLWordPrintable

    Details

    • Type: Story
    • Status: In Progress
    • Resolution: Unresolved
    • Fix Version/s: None
    • Component/s: None
    • Team:
      Data Release Production
    • Urgent?:
      No

      Description

      Discussion during cell-based coadd sprint week led to the decision to add a generic function to meas_algorithms to remove the Poisson contribution from actual sources to variance planes. This is done by estimating the gain through a linear fit.

      A similar thing is done in meas_extensions_piff, except a single constant value averaged over the whole image is kept.

        Attachments

          Issue Links

          relates to

          Story - Created by Jira Software - do not edit or delete. Issue type for a user story. DM-32410 Prototype scheme for providing background-only noise estimates to measurements

          • To Do

            Activity

            Hide
            Permalink
            kannawad Arun Kannawadi added a comment -

            Jogging my memory here as I assign this to Erfan Nourbakhsh

            • IIRC, we decided to use this a free-floating utility function because we were going to use it in the descwl_coadd package. However, it might also be nicer to have this as a Task, whose run method calls this utility function. This would be akin to ScaleVarianceTask and SubtractBackgroundTask in that, it can be run either on calexps or on cell-coadds. The various arguments would be  Config fields.
            • To make sure that we don't accidentally subtract the source contribution twice (once on single-visit images, and once on coadds), it'd be nicer (but extremely difficult with afw Exposure objects) to have the ability to bookkeep whether the variance plane has been subtracted once or not. This is surely a Jim Bosch question.
            • We need to have checks to ensure that the variance plane doesn't become negative, and handle it appropriately.
            • I'd be curious to see how much smoother the variance plane gets, especially on pure background pixels, and whether it is meaningful since we are subtracting it by the image which strongly correlates with the variance plane (since that's how we obtained it in the first place)
            Show
            kannawad Arun Kannawadi added a comment - Jogging my memory here as I assign this to Erfan Nourbakhsh IIRC, we decided to use this a free-floating utility function because we were going to use it in the descwl_coadd package. However, it might also be nicer to have this as a Task, whose run method calls this utility function. This would be akin to ScaleVarianceTask and SubtractBackgroundTask in that, it can be run either on calexps or on cell-coadds. The various arguments would be  Config fields. To make sure that we don't accidentally subtract the source contribution twice (once on single-visit images, and once on coadds), it'd be nicer (but extremely difficult with afw Exposure objects) to have the ability to bookkeep whether the variance plane has been subtracted once or not. This is surely a Jim Bosch  question. We need to have checks to ensure that the variance plane doesn't become negative, and handle it appropriately. I'd be curious to see how much smoother the variance plane gets, especially on pure background pixels, and whether it is meaningful since we are subtracting it by the image which strongly correlates with the variance plane (since that's how we obtained it in the first place)

              People

              Assignee:
              erfan Erfan Nourbakhsh
              Reporter:
              mschmitz Morgan Schmitz
              Watchers:
              Arun Kannawadi, Erfan Nourbakhsh, Morgan Schmitz
              Votes:
              0 Vote for this issue
              Watchers:
              3 Start watching this issue

                Dates

                Created:
                Updated:

                  Jenkins

                  No builds found.