| The primary aim of this research work is to study and implement Viola and Jones [3] based face detector, and provide possible improvement to the original algorithm. It also studies its possible integration with real time face recognition system. Recently, many new face detection algorithms have been published based on original Viola and Jones' face detector. Viola and Jones provided three novel approaches for constructing a rapid and robust real time face detection system. Using Haar-like features with unique integral image representation the computation of the features becomes extremely rapid. Using AdaBoost a strong non-linear classifier can be constructed, and only useful features out of large feature set can be extracted and used for face detection. Cascade of classifiers provides very positive solution for improvement of execution speed and also improves the accuracy of the whole system. By cascading classifiers, each stage is trained separately using different feature, in other word each stage focuses on particular feature. Using this technique a very rapid face detector can be created and its results are comparable with other state-of-the art technique. To understand the role of face detection system in real time face recognition system eigenfaces based real-time face recognition is evaluated. |
