Online learning objectionable image filter based on SVM

In this paper we propose an on-line learning system for objectionable image filtering. Firstly, the system applies a robust skin detector to generate skin mask image, then features of color, skin texture and shape are extracted. Secondly these features are inputted into an on-line incremental learning module, which derives from support vector machine. The most difference between this method and other online SVM is that the new algorithm preserves not only support vectors but also the cases with longest distance from the decision surface, because the more representative patterns are the farthest examples away from the hyper-plane. Our system is tested on about 70000 images download from the Internet. Experimental results demonstrate the good performance when compared with other on-line learning method.