采用Gabor-Hough变换的自适应滤波人眼定位

人眼定位是人脸识别方法的第一步。传统人眼定位方法需要将人脸裁切,情况单一,对于复杂背景、倾斜等情况,精度低,宽容度差。指出将改进的滤波器与改进的定位方法相结合,首先对原图像进行Gabor变换,寻找图像凹陷,凹陷区域通过边缘像素向量改进的Hough变换检测瞳孔,得到参考坐标,经过训练得到滤波器并在角度[-0.1,0,0.1]旋转两次得到3个相关滤波器,通过相关滤波器滤出3个目标,选择最小误差位置作为最终的目标。相比传统方法,改进方法避免了矩阵盲目计算,具备传统方法的普适性,学习过程使定位误差减小为0.0404,计算量减少为传统方法的1/5。该方法对不同光照、光照不均、不同表情、复杂背景、头部倾斜等情况有较好的鲁棒性,避免了单一方法定位不到和不准的情况。

Eye location based on adaptive filter using Gabor-Hough transform

Eye location is the first step of the face recognition.Traditional eye location methods need to cut the face,and the situation is single. For complex background or leaned images,the methods have low accuracy and poor tolerance. In this paper,the improved adaptive filter is combined with the traditional positioning methods. Firstly,the original image is transformed with the Gabor Transform to find the sag regions. In these regions the pupil is detected with edge pixel vector improved Hough Transform to gain the first reference coordinates. The filter is trained and rotated for two times to get three coherent filters,then the image goes though the coherent filters to find three target coordinates.Finally,the best target which has the minimum error is chosen as the final target. Compared with the traditional methods,the new method avoids the weakness of aimless computation and has the general applicability of the traditional method. At the same time,the adaptive filter makes the mean error reduce to 0.040 4,and computing pressure reduce to 1/5.The method has better robustness for different illumination,uneven illumination,different countenance,complex background images and leaned head images,thus avoiding the situation of locating no targets and wrong targets with single method.