基于肤色和几何关系的人眼快速定位方法

Method on quickly Locating Eyes Based on Skin-color and Geometrical Relationship

人眼检测是人脸检测与识别、驾驶员行为分析或MPEG4压缩中的关键技术,为了提高处理速度和检测结果的鲁棒性,提出了一种基于肤色检测和几何特征人眼快速定位方法;通过比较,选用YIQ空间和KL变换联合的方法检测肤色区域,然后运用面积阈值检测出备选人脸区域,并在备选人脸区域中通过灰度特征确定人脸特征区域的位置,并根据人眼的几何位置关系检测出其大致位置;通过对称性和相似性校验所得位置是否为真正的人眼位置,最后运用Hough变换确定人眼瞳孔中心的精确位置;试验表明,该方法操作简单,速度较快,能满足实时处理的要求,对不同的光照条件、姿态以及干扰背景具有较强的适应性。

Eye detection is a crucial aspect in many useful applications ranging from face recognition/detection to human computer interface, driver behavior analysis, or compression techniques like MPEG,t. This paper presents a robust and efficient eye detection and location algorithm for color images based on skin color and geometrical relationship of facial features. By comparison of pixel classification perform ance, the mixture of YIQ color space and KL transformation is chosen to classify skin color pixels from background. Face candidates are ex traeted by the threshold of connection range. Then some blocks of low gray intensity arc got by threshold related with the mean gray inten sity of skin color pixels. Finally, eyes are located through the geometrical relationship of facial features. The detected eyes are verified by the symmetry and comparability, and the fine location of iris is confirmed by Hough transformation. According to the experimental results, the presented method can easy locate the eyes and is quick to meet the requirement of real-time detection. It can adapt to different illumination conditions, poses and complex background.