Figure 2From: A multi-cue spatio-temporal framework for automatic frontal face clustering in video sequences Movement dissimilarity between two non-overlapping tracks. Back to article page