摘要
文章在分析色情图像的特征的基础上,提出基于肤色模型与轮廓特征的图像过滤方法。重点讨论了基于内容的图像过滤方法,其中包括肤色检测、纹理检测、轮廓检测、特征选取和分类方法(Bayes分类器和SVM分类器)等关键技术。实验结果表明,在混合样本的条件下该方法能够达到80%以上的准确率。
The characters of the nude images are analyzed and an image filtering method is introduced,which is based on skin color model.The paper focuses on the method of content-based image classification,and discusses in detail related key techniques,such as skin color detection,skin texture detection,edge detection,feature extraction and classification methods （Bayes classification and SVM classification）.The experiments show that 80% images contain nude could be filtered accurately.
出处
《计算机工程与应用》
CSCD
北大核心
2006年第24期56-58,156,共4页
Computer Engineering and Applications
基金
四川省应用基础研究计划课题(编号:04JY029-096)
关键词
图像过滤
图像分类
肤色模型
数字图像处理
image filtering,image classification, skin color model,digital image processing
