An approach for improving face recognition in presence of inaccurate detection

In this paper, we introduce a new facerecognitionapproach robust to allocation error of face features. We show that combining a “standard method” for facerecognition and a new approach based on a “cloud of points” for representing a face image, we obtain a system that not only gives good performance when faces are perfectly detected, but also in the presence of detection errors. Extensive experiments carried out on the ORL and YALE databases of faces prove the advantages of the proposed approach when compared with other well-known techniques.