基于AdaBoost的肤色检测新方法

本文针对肤色检测问题，提出了一种利用AdaBoost方法构造分类器进行肤色检测的算法。根据肤色在色度空间内的聚类性，通过大量肤色和非肤色样本将一族弱学习算法通过一定规则训练成一个强学习算法，得到一个检测性能优异的肤色检测分类器。提出了用圆形分类器作为弱分类器描述色度空间中的肤色分布，将AdaBoost学习算法用于肤色的聚类分析中。实验表明，该方法误检率低、鲁棒性好，对肤色检测问题有较强的实用性。

A Novel Skin-Tone Detection Method Based on AdaBoost

Skin-tone color is a powerful fundamental cue that can be used at an early stage to detect faces in complex scene images. This paper presents a method for detecting skin-tone regions in color images based on the AdaBoost algorithm. Based on the clustering property of skin-tone color distribution in the chrominance space, we adopt a set of circle classifiers as simple "rules" which are trained via AdaBoost to form a strengthened classifier. The strengthened classifier has an outstanding ability for describing the distribution of skin-tone color in the YCrCb space. The results in this paper show that our method can provide robust detection with a low error rate and is applicable in the application of skin-tone detection.