期刊名称：International Journal of Signal Processing, Image Processing and Pattern Recognition
印刷版ISSN：2005-4254
出版年度：2015
卷号：8
期号：2
页码：227-234
DOI：10.14257/ijsip.2015.8.2.22
出版社：SERSC
摘要：To improve face detection rate and reduce false acceptance number with complex background, a method of face automatic detection based on improved elliptic skin extraction combining Adaboost algorithm is proposed. Firstly, the image is transformed from RGB space into nonlinear YCbCr space using nonlinear color transformation technology. Secondly, skin area is extracted based on elliptic skin model and after morphological image processing and face candidate region judgment, possible face region is determined preliminarily. Finally, face is detected accurately using improved cascade classifier. Experiments show that improved elliptic skin model can eliminate the influence of illumination and has good color extraction effects; face detection rate of proposed cascade classifier can reach 98% which is better than conventional algorithm. So the proposed method can enhance the performance and speed of face detection, and can detect face regions quickly and accurately with complex background
关键词：Pattern recognition; Face detection; Color space; Skin color model; ; Cascade classifier
