复杂背景下的多姿态人眼定位

针对彩色图像中人脸在复杂背景及多姿态下眼睛定位困难的情况,提出了一种基于肤色分割与Gabor滤波的人眼定位方法。首先采用粒子群算法优化改进的最大类间方差(Otsu)对图像进行肤色分割,找到人脸的候选区域;然后构造Gabor滤波器对候选区域进行滤波;最后经过灰度投影得到眼睛的精确位置。仿真结果表明,该分割算法在复杂背景和多姿态情况下的人眼定位中有明显优势,对于现实环境中的人脸识别有重要意义。

Multi-position eye location based on complex background

To solve the problem of eye location of human face in multi-position and complex background of color image, this paper proposed a method for eye location based on skin segmentation and Gabor filter. Firstly, it used the improved Otsu algorithm based on particle swarm optimization to segment the image and obtained candidate regions of face. Then it devised an Gabor filter to filter the image to highlight the eye region. Finally, it used the gray projection to determine the position of eyes accurately. The results show that the method of skin segmentation works well and has advantages about eye location of human face in multi-position and complex background, which is important for face recognition in real time.