Adult image detection based on skin visual words and face information

Adult image detection is an important technology for Internet pornographic information filtering. However, the conventional methods, such as analyzing skin-like or global features, suffer from the high false positive rate. Based on the Bag-of-Words model, a new adult image detection method combined with skin visual words and face information is proposed. Firstly, a skin color detection algorithm based on wavelet transform is used to estimate the skin areas. If skin ration is too small, the image is normal. Secondly, local descriptors are extracted from the interest points in skin areas to construct the skin visual words, and then the image is described as a vector of skin visual words occurrences to be classified by the SVM classifier. Finally the face information based on the deformable part model is used to eliminate the misjudged images with large face areas. The experiment results show that the new method can reduce the false positive rate effectively on the premise of the true positive rate unchanged, compared with the state-of-the-art methods.