提出了一种基于曲线波隐马尔可夫模型的混合人脸检测算法。曲线波变换是一种多尺度几何变换,具有很好的方向性,能用极少的非零系数精确表示图像的边缘,是一种最稀疏的表示方法。根据隐马尔可夫模型对人脸拓扑结构的约束,采用3状态的隐马尔可夫模型进行从粗到细的人脸检测。实验结果表明,这种算法具有较好的检测速度与正确率及鲁棒性。
A method of face detection based on Curvelet Hidden Markov Model is presented for this paper.Curvelet transform is a kind of multiscale transform,it has a good property of directivity and describes image edge through very few coefficients,so Curvelet is a type of the most sparse representation way of image.According to restriction of face topology by Hidden Markov Model,the 3-State Curvelet Hidden Markov Model is used to detect face from coarse to fine.Experimental results show this method has good speed,considerably correctness and robustness in face detection.