Face recognition from a single sample per person based on LTP

Face recognition from a single sample per person has been an active research area in the past two decades with a lot of encouraging results reported. For face recognition from a single sample per person, a generic learning method based on LTP is proposed. In order to make the database “generic” as well as reasonably sized, the whole data set is collected from 3 well-known databases, FERET, BioID and CAS-PEAL. The normalized method and image processing method are introduced to process the generic training set. Different sub-sets are divided and multi-pass recognition scheme is introduced. Experimental results on FERET, BioID and CAS-PEAL databases demonstrate the effectiveness of our generic learning method when the illumination is changeable.