自适应分数阶微分在面部图像处理中的应用

摘要

针对三维人脸表情识别过程中,皱纹等纹理特征会影响表情识别的精度.为了消除这类纹理特征的干扰,需要先确定皱纹的位置.因此,提出了一种基于自适应分数阶微分掩模的皱纹检测方法来定位皱纹的位置.该方法利用皱纹的纹理方向信息,构建自相关微分掩模区域,并使用皱纹图像的梯度、信息熵和对比度信息来改进分数阶微分掩模算子,对人脸皱纹进行纹理方向上的自适应增强.然后结合Gabor滤波器来提取皱纹曲线位置,再根据定义的误判原则剔除光照、皮肤色斑等误识别特征.实验结果表明,提出的方法能自适应增强皱纹深浅不一的面部图像,提升了对浅皱纹的检测率,同时借助误识别过滤规则,提高了皱纹位置检测的精度和鲁棒性.%In the process of 3D facial expression recognition,wrinkles and other skin surface texture would affect its accuracy.To localize the facial wrinkles and eliminate the interference,a new wrinkle detecting means is proposed based on adaptive fractional differential mask.An autocorrelation differential mask region is constructed according to the direction of wrinkles.And the fractional differential mask operator,which is improved by image gradient,entropy and contrast,is utilized for enhancing the facial wrinkles adaptively.Then Gabor filter is used to obtain curvilinear shapes of wrinkles and the discrimination rules are defined to minimize the false positive caused by light and skin discolorations.Experimental results show that the proposed method can be used to enhance the facial wrinkles of various depths adaptively,thus improving the detection of light wrinkles and the accuracy and robustness of wrinkle detection through the discrimination rules.