提出了一种人脸表情的矢量分解与合成算法,能将任意图像分解为6种基本表情的矢量和。根据分解的结果,可以实现表情强度的估计和表情的识别,也可以根据给定的表情强度参数合成各种表情图像,克服了表情强度估计、表情的分解与合成中需要手工标注特征点的问题,降低了训练集的不同对表情识别的影响。
A vector resolution and composition algorithm of facial expression is proposed. An expression image can be decomposed into vector addition of six basic expressions using the algorithm. Expression intensity is estimated according to the decomposed results. The algorithm can also synthesize the expression image in light of given expression intensity values. The algorithm settles the problem that feature points had to be labeled by hand in lots of expression intensity estimation, facial expression decomposition and composition algorithms, and reduces the influence of training set on expression recognition.