一种改进的基于几何特征的人眼定位方法

针对灰度图像,提出一种改进的基于人眼几何特征的实现人眼快速定位的方法.先对人脸进行预处理,再根据人眼在二值图像中的几何位置特征制定人眼判定规则并粗估计图像分割的灰度门限所在范围,在该范围内通过递增灰度门限的方法,在不同灰度门限下的人脸二值图像中寻找符合规则的人眼黑块,最后通过二维相关系数来验证.抽取ORL人脸图库中24人的图片进行实验,结果比未经预处理的图片准确率提高了15%,从而证明这种算法是行之有效的.

An improved geometry-based eye location method

In allusion to grayscale images,an improved geomertry-based eye location method was proposed.A range of threshold values that would separate eye blocks from others in a segmented facial image was estimated.With the progressive increased of the threshold by an appropriate step in that range,the gray image after preprocessing was converted into binary image.Labeled each black block in the binary image and checked them according to the determination criterions established by the priori knowledge of geometrical facial features,looked for the potential eye blocks.Once two potential eyes blocks appeared from the segmented image,a 2-D correlation coefficient was used as a symmetry similarity measure to check the reliability of detected eyes.Experimental results on ORL face database confirm that the proposed method lends itself to higher accuracy.