基于YCb＇Cr＇肤色特征和Otsu算法的人眼定位优化方法

为了解决光照和人脸姿态对人眼定位的影响,提出一种基于肤色特征和最大类间方差法（Otsu）相结合的人眼定位优化方法.该方法首先在检测出人脸的基础上,在YCb＇ Cr’色彩空间里做肤色检测,运用肤色分割原理缩小人眼检测的搜索区域;接着,对图像进行Otsu阈值分割,提取二值图像的眼睛区域;最后,在眉眼区域利用二值积分投影分别定位左右眼睛.该方法减小了光照对图像的影响,并且简单快速.在标准的VidTIMIT人脸库和自采图库上进行了实验仿真,结果表明,此算法在光照和人脸姿态变化的情况下都具有很好的鲁棒性.

Eye Localization Optimization Method Based on YCb＇ Cr＇ Skin Color Feature and Otsu Algorithm

To alleviate the impact of illumination and face-poses on eye localization,an eye localization optimization method is proposed based on the skin color feature and maximum inter-class variance （Otsu） algorithm.Firstly,following the detection of the human face,the skin color is detected in Ycb＇ Cr＇ color space.The skin color segmentation principle is used to narrow the search region in human eye detection.Then,the segmented image is converted to a binary image by Otsu algorithm and the eye region is extracted.Finally,two eyes are positioned using binary integral projection.The proposed method is simple and fast and reduces the impact of illumination on the image.Experimental simulation is done using the standard VidTIMIT face database and self-collected image database.Simulation results reveal that the algorithm is robust against changes of illumination and face-pose.