一种新颖鲁棒的嘴唇轮廓提取方法

Novel and Robust Approach to Extraction Outlines of Lip

嘴唇是人脸特征的重要组成部分,在音视频识别、认证、嘴唇同步、人脸识别等方面都很重要.但因嘴唇与肤色间对比度差,使嘴唇分割十分困难.根据在RGB颜色空间中,红色分量与绿色分量间的差异在嘴唇处较人脸其他肤色处明显,利用伪色调,结合YIQ彩色空间中Q分量在嘴唇处较稳定这一特点,首先确定出嘴唇上下边界.利用人眼与其他肤色存在明显差异,确定嘴唇的左右边界.然后对嘴唇区域进行形态操作、边缘提取等以得到完整的嘴唇轮廓.通过对大量不同肤色的人脸图像的嘴唇轮廓提取,证实了该方法的有效性和鲁棒性.

Lip is one of the most important facial features, and plays key roles in a variety of fields such as audio-visual speech, authentication, lip synchronization, face recognition, etc. The lip segmentation, however, is very difficult due to lower contrast between the lip and skin. According to the fact that the difference between the R and G components in RGB color space is greater for lip than for skin, the upper and lower boundaries of lip can be determined at first by pseudo-hue and utilizing Q component of YIQ color space in lip with more stable characteristics. Utilizing the characteristics of eyes with obvious difference from other skin, we locate the left and right borders of the lip. The complete outlines of the lip are extracted by a series of operation such as morphologic analysis, edge detection, etc. across the segmented lip region. Experimental results on extraction of the lip Outlines across a number of face images with different skins have demonstrated that our developed algorithm is effective and robust.