CV Reduce Values By Label

Collapses a labeled value array (e.g. the per-point angles of 'CV Points To Polar' + the labels of the 'CV K-Means Points' blueprint) to one number per cluster: the mean, the median, or the mode - the center of the densest window of the cluster's value distribution. The mode is the robust choice when a minority of contaminating members sits to one side (a numeral fused to a clock hand shifts the mean AND the median, but the hand's collinear pixels still form the densest spike). Rows follow the label order (0, 1, ...), aligned with 'CV Reduce Points By Label' over the same labels. Empty input yields empty outputs. For an IMAGE-scale reduction (multi-channel, a 2-D label map, rows indexed by label value) use 'CV Reduce Array By Label' instead - this one is a per-POINT reducer and its mode path is quadratic in the cluster size.