Description
Pull Request #2503 (for Issue #2482) fixes a bug where the calculated median was quite wrong in some situations. The precise calculation that is now used can use as much additional memory as the input image (if most pixels have the same non-zero label) and sometimes this input image could be large. This is a request that an option be added, e.g. SetConserveMemory, for the itk::StatisticsLabelMapFilter class (and, as appropriate, for its subclasses) and a supporting implementation be created that computes the precise median more efficiently memory-wise (though less efficiently time-wise).
Description
Pull Request #2503 (for Issue #2482) fixes a bug where the calculated median was quite wrong in some situations. The precise calculation that is now used can use as much additional memory as the input image (if most pixels have the same non-zero label) and sometimes this input image could be large. This is a request that an option be added, e.g.
SetConserveMemory, for theitk::StatisticsLabelMapFilterclass (and, as appropriate, for its subclasses) and a supporting implementation be created that computes the precise median more efficiently memory-wise (though less efficiently time-wise).