<XML><RECORDS><RECORD><REFERENCE_TYPE>1</REFERENCE_TYPE><REFNUM>7764</REFNUM><AUTHORS><AUTHOR>Balasuriya,L.S.</AUTHOR></AUTHORS><YEAR>2000</YEAR><TITLE>Frontal View Human Face Detection and Recognition</TITLE><PLACE_PUBLISHED>B.Sc. Honours thesis, Department of Statistics and Computer Science, University of Colombo </PLACE_PUBLISHED><PUBLISHER>N/A</PUBLISHER><LABEL>Balasuriya:2000:7764</LABEL><KEYWORDS><KEYWORD>Face recognition</KEYWORD></KEYWORDS<ABSTRACT>This thesis is an attempt to unravel the classical problem of human face recognition. The researcher addressed the problem of automated face recognition by functionally dividing it into face detection and face recognition. Different approaches to the problems of face detection and face recognition were evaluated, and five systems were proposed and implemented using the Matlab technical computing language. In the implemented frontal-view face detection systems, automated face detection was achieved using a deformable template algorithm based on image invariants. The deformable template was implemented with a perceptron. Unsupervised learning using Kohonen Feature Maps was used to create the Perceptron's A-units. The natural symmetry of faces was utilised to improve the efficiency of the face detection model. The deformable template was run down the line of symmetry of the face in search of the exact face location. Automated frontal view face recognition was realised using Principal Component Analysis, also known as the Karhunen-Loeve transform. Manual face detection was used to test the implemented automated face recognition system. The frontal view face recognition system is also expanded into a pose invariant face recognition system which is implemented and tested on facial images for subjects with different poses. The researcher gathered a face database of 30 individuals consisting of over 450 facial images to test fully automated face detection without verification, fully automated face detection with verification, manual face detection and automated face recognition, fully automated face detection and recognition and pose invariant face recognition. Successful results were obtained for automated face detection and for automated face recognition under robust conditions. Fully automated face detection and recognition was not realised because an eye detection system could not be implemented. Pose invariant face recognition was also successfully implemented under controlled conditions. </ABSTRACT></RECORD></RECORDS></XML>