Multiple media based face recognition in unconstrained environments using eigenfaces

Face recognition is a challenging problem in computer vision and image processing domain and it is because of its numerous applications including video surveillance, Remote System logon, and Passport verification and so on. Among them face recognition is most important in Law Enforcement and Security area because of its unconstrained nature. Unconstrained Face Recognition differs from normal face recognition by the way it deals with the images and it uses images that are captured without target's co-operation. In a city crime including robbery, murder etc. the available sources to find the suspect are images or videos taken by the common people using their cell phones, Surveillance video frames at the public places or some sort of verbal descriptions provided by the eye witness. Using these verbal descriptions a photo sketch can be developed and can be used as a source of information regarding the suspect. So, the survey on Unconstrained Face Recognition reveals that all the available image sources can contribute a better identity to the target suspect. Thus this paper aims to train an automatic face recognition system that takes multiple media sources as inputs to recognize a person of interest from gallery of images using Principle Component Analysis.