Skip to content

Hpd image fitting s-curve normalisation

Christopher Rob Jones requested to merge HPDImageFitting-SCurveNormalisation-2016 into 2016-patches

More improvements to the HPD image fitting used to create the HPD image calibration constants.

  • More improvements to the memory management. Include in the returned fit result object shared pointers to the various histograms used during the fitting procedure. Removes the need for callers to recreate them if required, as now they can just directly access those used internally by the fit itself.

  • Add a new processing mode, which normalises the histograms using a SCurve (tanh) method, effectively removing areas of very high intensity, which can cause the Sobel filter some problems. Better approach than the old (although still there) Log(z) as it does not effect the extracted radius in the same way, and is more accurate. The returned errors are also much more realistic.

No impact on the reconstruction, either HLT or offline, is expected as this algorithm is not used as part of the reconstruction sequence.

Merge request reports