复杂背景正面人脸嘴唇检测算法研究

Locating algorithm of front face's lip in complex background

嘴唇是人脸特征的重要组成部分,嘴部的定位和识别在口型识别与跟踪,唇读以及人脸动画合成等方面都起着十分关键的作用。首先在YCbCr空间建立肤色模型,对复杂背景图像中的人脸区域进行检测、定位,并进行去噪处理;然后在标准RGB彩色空间中,利用唇色进行嘴唇区域的提取。实验结果表明该方法可以完成在不同光照、不同背景下的人脸嘴唇区域的快速检测和定位。

Lip is one of the most important facial features. Lip localization and recognition plays key roles in a variety of fields such as embouchure identification and tracking, lipreading, facial animation synthesis, etc. Firstly, a skin-color model is constructed in YCbCr color space so that human face can be detection and localization in complex background, and the noises of image will be removed. Then, lip outline can be extracted from RGB color space by lip color. Experimental results show that the proposed method is effective in front face＇s lip rapid detection and localization in different light and background.