Faceprint

This thesis presents a comprehensive overview of the problem of facial recognition. A survey of available facial recognition algorithms as well as implementation and tests of a computationally efficient and near real time well established approach to face recognition is presented. One of the oldest and robust face recognition algorithms, Eigenface, along with enhancement to that algorithm Histogram Equalization, Circular Tracing, Principal Component Analysis and Averaging techniques are added to get better results to identify individual from face dataset. A proof of concept implementation is provided, that is written in Visual Basic 6 on Visual Studio 6 platform. Extensive testing of the project along with performance measures are presented in easy to understand graphical images. These test evaluations showed that the proposed implementation of eigenface technique can be increased by better preprocessing and normalization of the input face space before raw eigenface approach takes over. Tests also suggested that the core eigenface technique hits a plateau once it hits the threshold of optimal number of eigenvectors. Finally, this thesis presents a discussion of the test results and a section on future direction this project may be led on.