For the purpose of the real-time face detection system, we used to image from the camera. We use a training algorithm of Adaptive Boosting based on the three type of Haar-wavelet features. Adaptive Boosting algorithm's defect is that it takes very lon...
For the purpose of the real-time face detection system, we used to image from the camera. We use a training algorithm of Adaptive Boosting based on the three type of Haar-wavelet features. Adaptive Boosting algorithm's defect is that it takes very long time and even if one of training data changes, retraining is needed. So we have proposed a training system of lower training time and robust to changing the training data. The training procedure is separated according to each level, so informative date applied to next stage recursively. Therefore, we have created strong classifier which has robust detection rates using lower training time and data. Experimental results show that our system is robust to various face and illumination change at real-time image data.