Unified Probabilistic Models for Face Recognition from a SingleExample Image per Person

This paper presents a new technique of unified probabilistic models for facerecognition from only one single example image per person. The unifiedmodels, trained on an obtained training set with multiple samples perperson, are used to recognize facial images from another disjointdatabase with a single sample per person. Variations between facialimages are modeled as two unified probabilistic models: within-classvariations and between-class variations. Gaussian MixtureModels are used to approximate the distributions of the two variations andexploit a classifier combination method to improve the performance.Extensive experimental results on the ORL face database and the authors'database (the ICT-JDL database) including totally 1,750 facial images of350 individuals demonstrate that the proposed technique, compared withtraditional eigenface method and some well-known traditionalalgorithms, is a significantly more effective and robust approach for facerecognition.