基于改进对称变换的人眼定位

Eyes location based on improved symmetry transform

目的提出一种定位人脸图像中眼睛的方法。方法应用广义对称变换的原理,对传统的对称点对区域的计算公式进行了改进,采用基于面积系数的圆环邻域模板,取代了传统方法的对称算子圆形邻域,该法引入了更多边缘像素的梯度信息,保持了边缘像素之间的连续性。结果通过对传统算法进行改进,有效地提高了眼睛定位的精度和鲁棒性。结论实验表明算法有效,具有一定的实用性。

Aim To properse a method of eyes location on human face image. Methods Based on generalized symmetry transform principle, an improved strategy for computing symmetry pixels-pairs neighbor zone is proposed, the traditional circle mask of symmetry operator is replaced by area coefficient based annular mask, which can pro- vide more gradient information about objects＇ boundaries pixels, and keeps the boundaries pixels continuous. Results Experiments results show that the algorithm is efficient and practical. Conclusion The performance of accuracy and robust can be obtained by improving traditional algorithm.