Using Skin Color and HAD-AdaBoost Algorithm for Face Detection in Color Images

Owing to the interference of the complex background in color image, high false positive rate is a problem in face detection based on AdaBoost algorithm. In addition, the training process of AdaBoost is very time consuming. To address these problems, this paper proposes a two-stage face detection method using skin color segmentation and heuristics-structured adaptive to detection AdaBoost (HAD-AdaBoost) algorithm. Firstly, skin color segmentation is applied to remove most of the background. Then, the HAD-AdaBoost cascade classifier is performed on the candidates to make a final face location. By introducing the algorithm of adaptive to detection AdaBoost algorithm (AD AdaBoost) and heuristicsstructured cascade to our system, the resulting classifier consisting of fewer weak classifiers achieves lower error rates. The experimental results demonstrate that our system retrieves 90.24% of the detection rate with 1.69% false alarms on the color image set. Keywords-face detection; skin color segmentation; AD AdaBoost; heuristics-structured cascade