Abstract: A lip detection method is put forward, according to the difference properties between face color and lip color in R, G, B chromaticity component on the distribution between the yellow race and the white race. This method identifies the threshold by the chromaticity difference, and segments the lip rectangular area by using projection algorithm, then extracts the lip combined with the edge detection and dynamic threshold method. Compared with the Red Exclusion algorithm, the results show that, the algorithm is simple, quick, and the accuracy of this detection algorithm is improved and advanced such as different shades of color, uneven illumination and a certain degree of angular deflection of the head etc.

Key words: projection algorithm, dynamic threshold, edge detection, extraction

摘要： 根据黄色人种和白色人种的肤色和唇色在R、G、B色度分量上分布的差异，提出了一种改进式的唇部检测方法。该方法根据色度差异的特性确定阈值，利用投影算法分割出嘴唇所在的矩形区域，结合边缘检测和动态阈值的方法提取出嘴唇。与红色排除法相比，结果表明，该算法简单、快捷，对肤色深浅不同、光照不均匀及头部有一定的偏转等情况，该算法检测的准确性都有一定的提高和改进。

关键词: 投影算法, 动态阈值, 边缘检测, 提取