Inconsistent no-data checks in StreamingVectorImageFilter<> and StreamingHistogramVectorImageFilter<>.
Mantis Issue 919, reported by salbert, assigned to msavinaud, created: 2014-03-27
StreamingHistogramVectorImageFilter.txx:257 tests (vector) PixelType[ band ] against no-data value (because it creates a single one-dimension histogram for each band) i.e. pixel[ b ]==noData. So, it rejects all pixels with [NoData, val1, val2, ...] and all permutations.
StreamingStatisticsVectorImageFitler tests InternalPixelType against no-data value to compute min and max pixels i.e. internalPixel==noData. So it rejects only the pixel [NoData, NoData, NoData ...].
So, using StatisticsVIF to get min/max pixels and using HistogramVIF to compute each band (one-dimensional) histogram leads to numerical issues.
E.g. (use of Monteverdi2) min pixel is 0 and itk::Histogram<>::Quantile(0, 0) returns NaN.
1396002570 - mickaelfix with http://hg.orfeo-toolbox.org/Monteverdi2/rev/b26f1196559a
1429262748 - julienClosing resolved bugs.