摘要
Adaboost算法是一种被广泛应用于人脸检测的分类器学习方法,通过Haar-like特征和样本的学习和训练,形成一个强分类器,能有效地区分人脸跟非人脸。文中提出一种Adaboost结合最小割算法的人脸提取方法,该方法着眼于图像中的轮廓及肤色信息,对每个点设置一个权值,寻找一条权值最小的边界,准确提取出人脸。实验结果表明,Adaboost和最小割的人脸提取算法,分割效果较好,且耗时较小。
Adaboost algorithm is widely used in face detection. It＇s a face classify learning algorithm method. Through the learning over Haar-like features and samples, it develops a strong classifier that can divide face and non-face region. Arrange the strong classify in some order, which forms a more efficient face classify. In this paper, we pro- pose an algorithm based on rain-cut and Adaboost algorithm in order to detect face better. This algorithm focuses on the outline and the skin-color pixel in the image, and finds the boundary with least total cost. The experiment result showed this algorithm can achieve face division efficiently.
出处
《自动化与仪表》
北大核心
2010年第9期52-55,共4页
Automation & Instrumentation
基金
国家自然科学基金项目(60872123)
国家自然科学基金重点项目资助(U0835001)
