摘要
人脸检测是自动人脸识别中的关键环节。提出了基于人脸肤色模型和人脸结构特征的人脸检测。首先利用人脸图片样本,提取肤色像素,建立肤色CbCr高斯模型。根据高斯肤色模型求得人脸似然图,并采用最佳阈值对之进行二值化。之后再采取形态学处理,除去部分非人脸区域,分割出待定人脸区域。最后根据人脸的结构特征进行再次筛选,得到人脸区域。此方法运用在一般的人物照片中都能达到比较理想的效果。
Face detection plays a key role in automatic face recognition.A human face detection method based on skin color model and structural features of face is presented.Firstly,skin color pixels are extracted from face images,and CbCr Gaussian model can be constructed.Secondly,the likely region of face is gained according to the Gaussian model,and the binary image can be got by the best threshold.And then,use morphological processing to remove some non-face part of the image,so can be got the suspicious face region.Finally,the face region is chosen according to the structural features of face.The ideal effect can be achieved through using the method in the general character photos.
出处
《科学技术与工程》
2010年第21期5179-5183,共5页
Science Technology and Engineering
基金
中国博士后科学基金(20080440923)资助
关键词
人脸检测
肤色模型
结构特征
face detection skin color model structural features
