A New Face Detection Method Based on Improved Adaboost and Skin Color Model in Complex Background

摘要

Two improvements of Adaboost based face detection algorithm are provided in this paper. Firstly, dual-threshold weak classifiers are used to improve the detection accuracy. Secondly, Pearson's correlation coefficient is used as a measurement of correlation between classifiers, which significantly reduces the redundancy features of classic Adaboost algorithm. Through these improvements the detection speed is accelerated significantly. Finally, a new face detection method combining skin color model and improved AdaBoost algorithm is proposed, which is an efficient face detection method with high accuracy. Also some experiments are made to certify these improvements.