Uploaded image for project: 'Data Management'
  1. Data Management
  2. DM-28928

Prototype image split code for spatially varying zogy and decorrelation afterburner

    XMLWordPrintable

    Details

    • Type: Story
    • Status: Done
    • Resolution: Done
    • Fix Version/s: None
    • Component/s: None
    • Labels:
      None
    • Story Points:
      6
    • Epic Link:
    • Sprint:
      AP S21-5 (April)
    • Team:
      Alert Production
    • Urgent?:
      No

      Description

      Create prototype code and demonstration notebook for the implementation of the spatially varying zogy and decorrelation afterburner implementation.  We have the following preferences:

      • Each processing sub-image consists of a useful inner box and a padding outer box. The calculation will be performed on the sub-image but only the inner box result is kept.
      • We make a conservative estimate of the zogy matching kernel/decorrelation kernel size in image space. The outer border should be at least 1 kernel size wide on each side (1/2 kernel size is the effect range of the kernel at the edge of the inner box + 1/2 kernel size to avoid mirroring due to the circular boundary condition).
      • The whole sub-image should be of power of 2 dimensions unless this is too large.
      • The inner box should be at least about ~10 matching kernel size, to better average out the high frequency noise that may remain in the kernel.
      • The inner box should not be small near the edges due to rounding.
      • The padding edge should be filled with random noise. Test how to do this efficiently - is it worth of using a pre-generated random noise buffer?
      • The padding edge can be divided into smaller boxes for fast determination of containment within the whole image. Only the sub boxes that are not fully contained should be pre-filled with random noise, the fully contained ones can be omitted (will be overwritten by the image data).

        Attachments

          Issue Links

            Activity

            Hide
            gkovacs Gabor Kovacs [X] (Inactive) added a comment - - edited

            Kenneth Herner Would you please review this ticket? I'm happy to explain the considerations why zogy/decorrelation needs this; otherwise the code is just a series of lsst.geom.Box2I generation and some interactive display of them.

            Show
            gkovacs Gabor Kovacs [X] (Inactive) added a comment - - edited Kenneth Herner Would you please review this ticket? I'm happy to explain the considerations why zogy/decorrelation needs this; otherwise the code is just a series of lsst.geom.Box2I generation and some interactive display of them.
            Hide
            kherner Kenneth Herner added a comment -

            Sure, I can review it but can't get to it today. I'll start on it tomorrow.

            Show
            kherner Kenneth Herner added a comment - Sure, I can review it but can't get to it today. I'll start on it tomorrow.
            Hide
            kherner Kenneth Herner added a comment -

            I finally made it through; the logic appears sound though I'm not exactly sure how the erodedBy and dilatedBy functions work. One other thing that was not clear to me if why you do 

             

            b = a.erodedBy(2)

            b.shift(Extent2I(1,0))

             

            twice, once in cell 3 and again in cell 4.

            Show
            kherner Kenneth Herner added a comment - I finally made it through; the logic appears sound though I'm not exactly sure how the erodedBy and dilatedBy functions work. One other thing that was not clear to me if why you do    b = a.erodedBy(2) b.shift(Extent2I(1,0))   twice, once in cell 3 and again in cell 4.
            Hide
            gkovacs Gabor Kovacs [X] (Inactive) added a comment -

            The duplication is just copy-paste cleanup mistake. Otherwise, I wanted to create an asymmetric case to test the coverage of the border area by the sub boxes.

            Show
            gkovacs Gabor Kovacs [X] (Inactive) added a comment - The duplication is just copy-paste cleanup mistake. Otherwise, I wanted to create an asymmetric case to test the coverage of the border area by the sub boxes.

              People

              Assignee:
              gkovacs Gabor Kovacs [X] (Inactive)
              Reporter:
              gkovacs Gabor Kovacs [X] (Inactive)
              Reviewers:
              Kenneth Herner
              Watchers:
              Gabor Kovacs [X] (Inactive), Kenneth Herner
              Votes:
              0 Vote for this issue
              Watchers:
              2 Start watching this issue

                Dates

                Created:
                Updated:
                Resolved:

                  CI Builds

                  No builds found.