一种新的人脸图像中眼睛定位方法

Novel Method of Eye Detection in Facial Images

提出了一种新的图像中人眼检测算法。在一幅已知人脸图像区域中,首先采用中值滤波除去噪声点,然后选取灰度变化最大的几块区域作为眉眼候选区域。最后利用SIFT方法在候选区域中最终定位眼睛的精确位置、尺度和方向。实验结果证明,本文所提出的算法具有准确率高和计算量小的优点。

A new algorithm for eye detection in facial images was proposed. First, in a known facial image, the noises were cleaned with Median filtering method. Then, several areas where the gray-scale gradient was largest were chosen to be the candidates for brow- and- eye region. Finally, the exact positions,scales and directions of eyes were located by using SIFT algorithm. The experimental results demonstrated that the proposed algorithm has the advantages of higher accuracy rate and smaller computational complexity.