In recent years considerable progress has been made in the area of face recognition. Through the development of techniques like Eigenfaces computers can now outperform humans in many face recognition tasks, particularly those in which large databases of faces must be searched.
Whilst these methods performs extremely well under constrained conditions, the problem of face recognition under gross variations remains largely unsolved. This thesis details the development of a real-time face recognition system aimed to operate in less constrained environments. The system is capable of single scale recognition with an accuracy of 94% at 2 frames-per-second. A description is given on the issues and problems faced during the development of this system with particular focus on the difficulties encountered in multi-scale recognition.
It is concluded that: Eigenfaces are an excellent basis for face recognition system, providing high recognition accuracy and moderate insensitivity to lighting variations; Eigenfaces are sensitive to scale reductions of less than 88% and rotations of more than 10 degrees. Hence it is essential that good scale and rotation normalization algorithms be applied prior to recognition.
An overview of leading-edge developments in face recognition is given and conclusions drawn on where future research should be focused.