Rapid face detection using boosted simple features

Abstract

In recent years, face detection has attracted considerable interest from researchers and consumer-driven industries such as camera manufacturers. This computer technology identifies and locates human faces in an image regardless of position and scale. There are many applications based on face detection systems such as biometrics in facial recognition system, video surveillance and recent digital cameras for autofocus.

This project serves to compile three key discussions. All of which utilize Matlab to realize the idea of popular face detection suggested by P. Viola and M. Jones.

Firstly, algorithms were implemented to extract the coordinates and values of four feature types based on the introduction of “Integral Image” which enables simple features to be calculated in a rapid manner. Secondly, codes were executed to obtain face discriminating thresholds in order to build weak classifiers. Finally, a strong classifier was built based on Adaptive Boosting method and a real image detector was built based on fixed and varying threshold scale of face images. These allow us to investigate the performance of important features and develop other feasible parameters that can make the computation more efficient on real world images.

The attempt to successfully integrate new parameters into the algorithm is still ongoing. Further improvements that can be made to the system in this project for future work are recommended.