Finding faces in cluttered still images with few examples

Abstract

Elastic graph matching and its extension bunch graph matching have proven to be among the best methods for face recognition and the interpretation of facial expressions. We here demonstrate for the first time that, in combination with a simple color template, it is also an excellent means for the localization of faces in cluttered still images. The system does not need extensive learning, all information is extracted from a handful of example face images.