复杂背景图像中彩色人脸的检测

Face detection in color images with complex background

提出了一种在复杂背景的图像中自动检测彩色人脸的方法。该方法首先在YCrCb和HSV色彩空间进行肤色和非肤色的分割,对检测到的肤色像素在CrCb空间中进行聚类,在每一聚类中心应用形态学算子除去一些较小的背景区域,然后进行区域合并形成候选人脸区域。在候选人脸区域内应用重复阈值法得到候选眼睛对,最后采用BP神经网络进行确认。实验结果表明这种方法在复杂背景的图像中检测人脸的正确率为90%。

This paper presents a novel method to detect human face automatically in color image with (complex) background. Firstly, YCrCb and HSVcolor spaces are used to extract the skin color regions. In CrCb space, the detected skin pixels are clustered into several centers. Then, the opening morphological operation is used to remove the small objects in the background area. Some rules are employed to select the candidate face regions according to the shape and size information. Finally, the face regions can be located by using BP to detect the appearance of two eyes in the candidate face regions. Experimental results show that the proposed approach is efficient.