Breaking and Fixing gCNR and Histogram Matching

Generalized contrast-to-noise ratio (gCNR) and histogram matching are methods developed to better measure the performance of modern adaptive beamformers. Both use histograms to make comparisons between methods more robust and transformation-resistant compared to traditional metrics. However, the standard implementation of gCNR using histograms is potentially manipulable, and histogram matching can be difficult to implement and interpret. To make gCNR more robust, we suggest the disclosing of binning information and additionally the use of either rank-ordered data or variable bin widths. These two implementations increase the consistency of the gCNR metric and reduce the likelihood that the ranking of beamformers changes. For histogram matching, we consider situations where the choice of matching can yield different outcomes.