Constructing Pornographic Images Detector based on naïve Bayesian classifier

Abstract

Abstract
Detection of pornographic images can effectively prevent pornographic images from spreading on the Internet. This research proposes a new approach of pornographic images detector. Naive Bayesian classifier is used by the proposed detector to identify potential pornographic images. Skin and non-skin color models are constructed and exploited by constructing a Bayesian decision rule based skin detector. Several features are extracted from the output of skin detector which forms the features vector. The naive Bayesian classifier is trained on these features for both porn and non-porn classes. An experiment used (136) images for training the pornographic images detector and (154) images for testing it. The pornographic images detector is evaluated by using sensitivity, precision, specificity and accuracy metrics. It achieves a detective rate of (91.48%) with (6.67%) false positive rate.