Fast recognition of multiple faces using MCM

In this paper we present the novel minimal classification method (MCM) that provides an elegant and computationally efficient approach to the problem of multiple face recognition. In the classification we use a special set of filters which yield low values of correlation with a given set of images and large values with other natural images. Such filters are called antifaces. The minimal classification method can efficiently recognize individual images from a large set of 2/sup K/ images using only k filter sets whereas other methods require much more classifiers. We use here antifaces, but the same approach can be implemented with other pattern recognition methods such as SVM or eigenfaces.