Karim Sabri / ctools

Commit 718635be47eed799021a997d165f8456c14429f3

Authored by Jürgen Knödlseder
1 parent a7737cd8

Avoid negative squared standard deviations in cssrcdetect (#3261)

Inline Side-by-side
Showing 4 changed files with 19 additions and 15 deletions
ChangeLog
  View file @718635b
... ... @@ -3,10 +3,10 @@
3 3 * Version 1.7.0 released
4 4 ========================
5 5  
  6 + Avoid negative squared standard deviations in cssrcdetect (#3261)
6 7 Add Interface Manual to User documentation (#3163)
7 8 Adapt show_pull_histogram.py to matplotlib > 3.1.0 (#3156)
8 9 Add On/Off analysis to ctlike reference manual (#3154)
9   - Avoid negative squared standard deviations in cssrcdetect
10 10 Update user manual for latest CTA IRF (#3075)
11 11 Fix log file header repetition in some cscripts (#3066)
12 12 Add Moon constraints to csviscube (#3065)
... ...
... ... @@ -292,9 +292,11 @@ Add ebingamma parameter and support for POW binning (#2881).
292 292 cssrcdetect - Detects sources in sky map
293 293 ----------------------------------------
294 294 The using of a smoothing kernel in the _map_moments() method can lead to
295   -negative squared standard deviations that lead to an exception when the
296   -square root is taken. A filter was introduced that sets all small values
297   -of mean and std to zero, avoiding thus the exception.
  295 +negative squared standard deviations that lead to an exception when the square
  296 +root is taken. To avoid negative values, the minimum pixel values for the
  297 +standard deviation map is computed and, if the minimum value is negative,
  298 +subtracted from the standard deviation map. Therefore, by definition, the
  299 +standard deviation map is non negative for all pixels (#3261).
298 300  
299 301  
300 302 cstsdist - Computes Test Statistic distribution
... ...
cscripts/cssrcdetect.py
  View file @718635b
... ... @@ -237,16 +237,17 @@ class cssrcdetect(ctools.cscript):
237 237 # Compute the squared of the standard deviation for each pixel
238 238 std = std - (mean*mean)
239 239  
240   - # Set all pixels with small number of events or small number of
241   - # squared standard deviation to zero. Such pixels may occur due to
242   - # the smoothing operation that uses a fast fourrier transform.
243   - for iy in range(std.ny()):
244   - offset = iy * std.nx()
245   - for ix in range(std.nx()):
246   - index = ix + offset
247   - if mean[index] < 1.0 or std[index] < 1.0:
248   - mean[index] = 0.0
249   - std[index] = 0.0
  240 + # Compute minimum value of standard deviation
  241 + std_offset = 0.0
  242 + for index in range(std.npix()):
  243 + if std[index] < std_offset:
  244 + std_offset = std[index]
  245 +
  246 + # If minimum value is negative then subtract minimum value to guarantee
  247 + # a non-negative map of standard deviations
  248 + if std_offset < 0.0:
  249 + for index in range(std.npix()):
  250 + std[index] -= std_offset
250 251  
251 252 # Compute standard deviation
252 253 std = std.sqrt()
... ...
doc/source/admin/release_history/1.7.rst
  View file @718635b
... ... @@ -17,9 +17,10 @@ In particular, this release provides:
17 17 Bug fixes
18 18 ---------
19 19  
  20 +* [`3261 <https://cta-redmine.irap.omp.eu/issues/3261>`_] -
  21 + Avoid negative squared standard deviations in :ref:`cssrcdetect`
20 22 * [`3156 <https://cta-redmine.irap.omp.eu/issues/3156>`_] -
21 23 Adapt show_pull_histogram.py to matplotlib > 3.1.0
22   -* Avoid negative squared standard deviations in :ref:`cssrcdetect`
23 24 * [`3066 <https://cta-redmine.irap.omp.eu/issues/3066>`_] -
24 25 Fix log file header repetition in some cscripts
25 26  
... ...