非均匀光照下的人脸表情多状态检测与分类

Eye and Mouth Contours and States Detection in Facial Images with Diverse Expressions under Nonlinear Illumination

提出了一个用改进的不易受光照影响的ASM(Illum ination Insensitive-Active Shape Model,简称I2-ASM方法)进行眼睛和嘴巴轮廓检测以及用组合方法进行眼睛和嘴巴状态分类的算法。对有丰富表情的人脸图像进行实验,结果表明,提出的算法即使在非均匀光照下也能取得较好的轮廓检测和状态分类结果。

An improved ASM method, which is named I^2-ASM （ Illumination Insensitive - Active Shape Model）, is proposed to detect contours of eyes and mouths; meanwhile, combined methods are carefully designed and applied to detect eye and mouth states. Experiments, which are done with hundreds of facial images with diverse expressions, turn out satisfactory results. Both the contour detection algorithm and the states detection algorithm are verified to be effective even if the illumination is nonlinear.