基于肤色和类Harr特征的人脸图像的人眼检测

人眼检测在表情识别和计算机视觉领域得到了广泛的关注和研究，但是在多数的人眼检测方法中，对于背景较复杂的图像，识别率急速下降，误检率急剧上升。经过研究，使用椭圆肤色模型预处理图像，分割出肤色区域和非肤色区域，检测算法只对肤色区域进行人眼检测，有效降低了复杂背景造成的高误检率。同时特征选取是决定检测算法识别率和误检率等性能标准的关键因素，选取类Harr特征训练Adaboost级联分类器，实验表明了类Harr特征的有效性。

Eye detection in color face image based on skin and Harr feature

Eye detection has been given wide attention and research in face expression recognition and computer vision field,but the recognition rate descends and the false positive rate increases sharply when processing images with complex background in most eye detection algorithms.This paper preprocesses the color face image using Ellipse skin model,and extracts skin area and non-skin area.The eye detection algorithm only concerns the skin area and because of this,the false positive rare descends.At the same time,the feature selection is very important for pattern classification algorithm,and this paper selects the Harr-like features to train the Adaboost classifier.Experiments have shown the validity.