Class lsst::afw::math::Statistics

class Statistics

A class to evaluate image statistics

The basic strategy is to construct a Statistics object from an Image and a statement of what we want to know. The desired results can then be returned using Statistics methods. A StatisticsControl object is used to pass parameters. The statistics currently implemented are listed in the enum Properties in Statistics.h.

 // sets NumSigclip (3.0), and NumIter (3) for clipping
 lsst::afw::math::StatisticsControl sctrl(3.0, 3);

 sctrl.setNumSigmaClip(4.0);            // reset number of standard deviations for N-sigma clipping
 sctrl.setNumIter(5);                   // reset number of iterations for N-sigma clipping
 sctrl.setAndMask(0x1);                 // ignore pixels with these mask bits set
 sctrl.setNanSafe(true);                // check for NaNs & Infs, a bit slower (default=true)

 lsst::afw::math::Statistics statobj =
     lsst::afw::math::makeStatistics(*img, afwMath::NPOINT |
                                           afwMath::MEAN | afwMath::MEANCLIP, sctrl);
 double const n = statobj.getValue(lsst::afw::math::NPOINT);
 std::pair<double, double> const mean =
                                  statobj.getResult(lsst::afw::math::MEAN); // Returns (value, error)
 double const meanError = statobj.getError(lsst::afw::math::MEAN);                // just the error

Note

Factory function: We used a helper function, makeStatistics, rather that the constructor directly so that the compiler could deduce the types cf. std::make_pair()

Note

Inputs: The class Statistics is templated, and makeStatistics() can take either: (1) an image, (2) a maskedImage, or (3) a std::vector<> Overloaded makeStatistics() functions then wrap what they were passed in Image/Mask-like classes and call the Statistics constructor.

Note

Clipping: The clipping is done iteratively with numSigmaClip and numIter specified in the StatisticsControl object. The first clip (ie. the first iteration) is performed at: median +/- numSigmaClip*IQ_TO_STDEV*IQR, where IQ_TO_STDEV=~0.74 is the conversion factor between the IQR and sigma for a Gaussian distribution. All subsequent iterations perform clips at mean +/- numSigmaClip*stdev.