提出了一种新型快速的在红外图像中进行眼睛定位的方法。该方法由两部分组成:首先,通过同态滤波增强图像的对比度,使用改进的迭代阈值算法和积分投影函数获取人脸,根据人眼和人脸的几何特征,获取大致的人眼区域;其次,通过自适应中值滤波算法和迭代阈值法获取最小人眼块,利用最小人眼块的质心和积分投影最终准确定位眼睛。实验表明:所提出的算法是快速和可靠的。
A novel and quick method for eyes location in an infrared image is proposed. This method consists of two steps. Firstly, homomorphic filter is used to enhance the contrast of image. The human face is obtained through improved iterative threshold algorithm and integral projection function. According to the eyes and face geometry characteristics, the eyes region is obtained roughly. Secondly, the smallest eyes-block is obtained by ranked-order adaptive median filtering(RAMF) method and iterative threshold algorithm. Finally, the eyes position is obtained accurately according to the centroid of the smallest eyes-block and integral projection function. Experimental results show that the proposed method is fast and reliable.