An adult image detection algorithm based on Bag-of-Visual-Words and text information

Adult image detection plays an important role in Internet pornographic information detection and filtering. By analyzing the shortcomings of existing pornographic image detection algorithms depending only on image content or keywords of text, a new adult image detection algorithm fusing image semantic features and image-correspondent text information is proposed. Based on Bag-of-Visual-Words model, image visual features such as texture and local shape are coalesced with text information extracted from image file name, file header or webpage. Then the SVM classifier is applied to accomplish the image classification. The detection performance of proposed algorithm is validated by experiments on a variety of images with normal and pornographic content.