Last modified 9 years ago Last modified on 04/08/2010 03:30:03 PM

SDQA Image Background-Check Stage


Assess whether the background level in a Science Exposure is constant across specified spatial scales. Following ISR processing, which corrects for the electronic bias level, photometric non-uniformity, and pixel scale variation, the background should vary only slowly and smoothly on all spatial scales. The statistic being computed here should be minimized when this condition is true.


From Policy

  • maxBkgdRmsDeviation Maximum RMS deviation allowed in the distribution of background levels, in units of electrons/s
  • imageBkgdStatistic Choice of SDQA image statistic that best characterizes the background

From Clipboard

  • scienceFpaExposureId
  • Set of imageBkgdStatistic values computed by ISR for each Science_Amp_Exposure


The outputs are put on the clipboard as a persistable SDQA metadata vector with key = "imageBkgdRatings", ultimately to be stored in the database. The vector has the following quantities:

  • Deviation of background from mean over CCD (SDQA metric = "ccdBkgdDeviation")
  • Deviation of background from mean over FPA (SDQA metric = "fpaBkgdDeviation")


  1. Bias level/structure has been removed successfully from each Science_Amp_Exposure, and have been normalized to unit gain.
  2. Flat-field correction has been applied successfully (which also corrects for field illumination)
  3. Fringes have been removed from affected bands (i, z) [not clear if this will be done for DC3b/PT1]
  4. imageBkgdStatistic determination excludes all pixel values flagged in the static bad pixel mask, or that are saturated
  5. Remaining sources of background are the following:
    1. Sky (including moonlight), scattered light
    2. Residual detector/camera artifacts (ghost images [,fringing?]) not corrected by ISR
    3. Large, diffuse astrophysical sources (extended galaxies and nebulae)

Processing Flow

  1. Read scienceFpaExposureId from clipboard.
  2. Read imageBkgdStatistic values for all Amps over the Science_Amp_Exposure FPA from clipboard.
  3. Determine for each CCD, and for the FPA, the mean and FWHM of the distribution about the mean of the Amp-level imageBkgdStatistic values. This is defined for instance w.r.t. the image mode in following equation [specific statistic for PT1 is imageClippedMean with 4-sigma, 3 reject cycles]:

Equation describing figure of merit for background flatness

  1. ignore failed or missing per-Amp imageBkgdStatistic values [how do I know this?]
  2. Call this ccdBkgdDeviation and fpaBkgdDeviation for the accumulations over each CCD and the full FPA, respectively.
  1. Put SDQA ratings on clipboard for database storage: [names are TBD].
  2. Raise QualityThreshold event if:
    • ccdBkgdDeviation > maxBkgdRmsDeviation, or
    • fpaBkgdDeviation > maxBkgdRmsDeviation


  1. The choice of metric for imageBkgdStatistic above had been imageMode, but a signa-clipped imageMean will be used for PT1 at least. This choice should be deferable to a policy file.
  2. Should any background variation (linear gradient, say) be fit and removed prior to accumulating the statistics? [Tentative answer: no, not for PT1]
  3. How do I know whether a given imageBkgdStatistic value (Amp level) is valid? Assume for now it will not be produced/persisted if the corresponding pipeline stage fails.
  4. N.B. these metadata are persisted to the clipboard, but do not as yet appear in the database schema.


  • TimStat.gif (4.7 KB) - added by shaw 9 years ago. Equation describing figure of merit for background flatness