摘要
人脸区域分割对于人脸信息处理研究有着非常重要的意义,既可以直接用于在图像中寻找人脸,又可以作为眼睛、嘴巴等人脸重要特征检测的前序研究。因此,提出了一种新的图像中实施人脸区域分割的算法。首先,利用同态滤波改善图像的光照条件并增强图像的对比度;然后,利用迭代式阈值选择算法对图像进行二值化并增强图像的对比度。在二值图像中依据一定的约束条件找到候选人脸并标定,然后搜寻二值图像中的"空洞"区域,如果"空洞"位置恰好呈倒锐角三角形分布,且位于候选人脸区域内,则证明此候选区域即为人脸区域。实验证明,该算法简单实用,而且很好地避免了光照影响,检测精度较高。
Face region segmentation plays an important role to the face information processing technologies. An algorithm to segment the face region is proposed. The homomorphic filter is used to improve the condition of illumination. The iterative thresh choosing algorithm is used to change the images into binary ones. According to some restrained conditions, the candidate field of the face is found. Then, the holes in the binary images are searched. If the locations of the holes distribute as a reverse acute-angled triangle and they lie in the candidate field of the face, the candidate filed is the real face region. Tests results show that, the algorithm is simple and efficient.
出处
《控制工程》
CSCD
2008年第5期587-589,共3页
Control Engineering of China
基金
教育部留学回国人员科研启动基金资助项目(20055383)
关键词
人脸识别
计算机视觉
轮廓
空洞
face recognition
computer vision
contour
holes
