Face Detection Using Asymmetric Boosting With Gabor Feature

With the continuous development of the digital image processing technology and intelligent learning algorithms, Face Detection technology is increasingly being applied to video surveillance, human-computer interaction and e-commerce and other fields. Face detection is a process of carving out the face objects from the static images or dynamic video frame background, and specifying the range of face area.Firstly, we summarize the existed face detection technology, and divide it into three categories:Feature-based face detection method, Based on template matching face detection method and Statistics-based face detection method. The paper points out that, at present the most widely used, the best accuracy and efficiency method is based on image features in statistics-based face detection method, and this paper introduces a core algorithm-AdaBoost algorithm.Secondly, a novel asymmetric Boosting algorithm is proposed for real-time face detection using improved Gabor features. The previous methods of asymmetric learning are mostly realized by the heuristic modification on the weights and confidence parameters of the discrete AdaBoost. In this paper, the asymmetric losses in cascade structure are explicitly optimized, and both the example weights and the classifier coefficients are learned in an asymmetric way.Thirdly, a kind of partially parallel architecture (PPA) is used to reduce both classifier training time and face detection time. Experimental results demonstrate that the proposed methods are better than the previous asymmetric Adaboost methods.