多层次特定类型图像过滤方法

A Hierarchical Method for Nude Image Filtering

随着网络和多媒体技术的发展 ,互联网中色情图像的传播愈演愈烈 ,为有效杜绝此现象 ,提出一种基于计算机视觉和模式识别的色情图像过滤方法——多层次特定类型图像过滤法 .该方法通过研究人类视觉对图像的分析机理 ,提出综合肤色模型检验、支持向量机 (SVM)分类和最近邻方法校验的多层次图像处理框架 .实验结果表明 ,该方法能够达到 85 %以上的准确率 .

Internet is a double edged technology, it pushes the development of multimedia, yet many pornographic images and videos might make problem to young generation. Techniques are needed to keep these images away from children. We developed a hierarchical nude image detection system. In the processing, images are first filtered by skin color model. Then, images are classified by support vector machine. Finally, images are verified by nearest neighbor. Experiments show that 85% images contain nude could be filtered accurately.