中文摘要：

针对传统Adaboost算法在光照不均匀、角度确不定的情况下，进行人脸检测出现检测率不佳的问题，文中基于Adaboost算法，提出一种改进的IB—Adaboost人脸检测方法。该方法通过对图像光照补偿预处理，改善图像光照质量；再基于Adaboost算法训练人脸分类器；接着对图像进行YCbCr色彩空间转换并二值化处理，缩小人脸搜索区域；再经过R、G、B颜色叠加获取皮肤区域的彩色图像实现人脸检测。实验结果表明，文中提出的IB—Adaboost方法在光照不均匀和人脸角度不定方面能够实现较为满意的检测效果。

英文摘要：

In view of the poor face detection performance of the traditional Adaboost algorithm under the condition of uneven illumination and uncertain angle of detection, this article puts forward an improved IB - Adaboost （ tiny illumination compensation and binarization） face detection methods. The method of image offers illumination compensation pretreatment to improve the image quality light; secondly, the face classifier is trained based on the Adaboost algorithm; then the YCb Cr color space transform and image binarization processing are performed to narrow the face search area; and finally the skin area color images are obtained by R, G, B color overlay to achieve face detec- tion. Experimental results show that the proposed IB -Adaboost method in uncertain uneven illumination and face Angle can achieve more satisfactory test results.