基于AdaBoost算法和色彩信息的脸部特征定位

LOCALISATION OF FACIAL FEATURES BASED ON ADABOOST ALGORITHM AND COLOUR INFORMATION

针对脸部特征定位提出一种新型的基于Ada Boost算法和色彩信息的方法。首先用参考白光照补偿法对脸部区域进行光线补偿,然后用Ada Boost算法的级联分类器快速定位图像中的脸部区域,最后根据脸部肤色与脸部特征在色彩信息上的区别,建立眼部模型和嘴部模型,实现眼睛与嘴巴定位。实验结果表明,基于该方法的脸部特征定位具有较高检测率和实用性,同时嘴巴定位不易受到表情变化的影响。

We proposed a novel method for facial features localisation,which is based on Ada Boost algorithm and colour information.First,we used the ＂reference white＂illumination compensation approach to compensate the light on face area; then we used the cascaded classifier of Ada Boost to rapidly locate the face area in image. Finally,according to the difference of colour information between complexion and facial features,we built eyes model and mouth model to realise the location of eyes and mouth. Experimental result showed that the facial feature localisation based on this method has higher detection rate and practicability,at the same time the mouth localisation is less conducive to be affected by face expressions as well.