Nude Image Detection Based on SVM

Abstract

On the Internet, the nude images caused the spread of a large number of social problems, how to identify the nude image accurately is a problem needing to be solved urgently. Therefore, we integrate both image processing method and support vector machines (SVM), this paper studies a new and enhanced approach on recognition of nude image, namely, combine a face detection model, skin color model and texture model, extract six nude image feature vectors. Additionally, some important factors of SVM are fixed by experiments, such as the training set, kernel function and the cost. The experimental results demonstrate that performing SVM-based nude image detective classification more effective in that it improves the prediction accuracies at the same time.