A New Effective System for Filtering Pornography Images from Web Pages and PDF Files

One of the more rapidly growing areas in search technology is image search. With this availability comes the natural need to filter offensive content, to prevent Pornography images from reaching the wrong eyes. Filtering and blocking software is one of the most frequently touted prevention devices. As any user of these services is aware, they often fail to remove offensive images. The reasons are clear in that “current Internet image search technology is based upon words, rather than image content”, as images are obtained by using the image filename or text that surrounds the image on a webpage. This paper presents an automatic software system for detecting and filtering Pornography images of any format (JPG, PNG, etc.) in web pages by using skin recognition. The proposed system is an online client-side filtering system that allows the user to choose skin detection filtering, keyword filtering, URL filtering, domain filtering, or combination of two or more of these filtering techniques. A hybrid skin color detection technique is proposed to overcome the failure of detecting complex image’s background. To the best of our knowledge, there is no tool to block pornographic images included in PDF documents. So, we augmented our proposed system with the capability of online and offline filtering of PDF documents based on skin color detection. The experimental results showed that our proposed filtering system is more effective and accurate compared to some of the established commercial filtering software.