[gnuastro-devel] [bug #53825] NoiseChisel not accounting for fully zero-valued tiles

[Mohammad Akhlaghi]

URL:
  <http://savannah.gnu.org/bugs/?53825>

                 Summary: NoiseChisel not accounting for fully zero-valued
tiles
                 Project: GNU Astronomy Utilities
            Submitted by: makhlaghi
            Submitted on: Thu 03 May 2018 08:07:06 PM CEST
                Category: NoiseChisel
                Severity: 3 - Normal
              Item Group: Crash
                  Status: In Progress
                 Privacy: Public
             Assigned to: makhlaghi
             Open/Closed: Open
         Discussion Lock: Any

    _______________________________________________________

Details:

It might happen that the input dataset has zero values for regions with no
data: essentially using zero as a mask value. In other detection methods (that
have above-Sky thresholds) this isn't an issue. But NoiseChisel's threshold is
below the Sky value, so the zero valued pixels will actually be considered as
detections. We therefore encourge users to set zero valued pixels to
NaN/blank.

When there are tiles with fully zero-valued pixels, they may thus be used for
estimating the pseudo-detection Sky and Sky STD and result in zero-valued Sky
(and its STD) values. This will later cause the counts-per-second correction
(`cpscorr') to be zero and produce `inf' S/N values.

It is thus necessary to also mask tiles that pass all other tests, but produce
a standard deviation of 0.




    _______________________________________________________

Reply to this item at:

  <http://savannah.gnu.org/bugs/?53825>

_______________________________________________
  Message sent via Savannah
  https://savannah.gnu.org/