Details
-
Type:
Story
-
Status: Done
-
Resolution: Done
-
Fix Version/s: None
-
Component/s: ip_isr
-
Labels:None
-
Story Points:8
-
Epic Link:
-
Sprint:DRP S19-6b
-
Team:Data Release Production
Description
At DRP Team meeting April 3 2019, we discussed the steps in ISR that could be reordered in order to support storing pre-interpolated pixels (that have had ISR applied to them so that they can be meaningfully swapped in downstream processing).
The interpolation step before brighter-fatter is necessary. Therefore, we decided to reorder ISRTask as follows:
- First ISR Steps…
- Interpolate
- Do Brighter-Fatter Correction
- Put the pre-interpolated pixels back into the now BF-corrected raw.
- …Continue ISR…
- Saturated/NaN pixels and defects are masked only (no interpolation)
- Second to last step: write out pre-interpolated pixels
- Last step: Interpolate the union pixels masked for interpolation (sat/bad/nan)
Robert Lupton pointed out that it is more correct to interpolate the union of all regions that need to be interpolated rather than in separate stages for each of SAT, BAD, NaN. In addition to the extra steps of reinserting pre-interp pixels, this tickets also includes refactoring to separate the masking step and interpolation steps in maskAndInterpNan, maskAndInterpDefects, and maybe saturationIntepolation.
Attachments
Issue Links
- is triggered by
-
DM-17471 Save intermediate data product with pre-interpolation pixel values
- Done
- mitigates
-
DM-18693 ip_isr will interpolate NaN values if brighter fatter is enabled
- Invalid
- relates to
-
DM-20342 Fix typo in eImageIsr
- Done
-
DM-20112 ip_isr is not handling BAD pixels as expected
- Done
-
DM-27319 Examine behavior of IsrTask wrt doSaturation and doSaturationInterpolation
- Done
Looks good; I left comments on the ip_isr PR and none of the others. Please add a template path for obs_decam though.