Uso de descritores binários para detecção de pornografia

Abstract

With the growing of the amount of inappropriate content on the Internet, such as pornography, it arises the need to detect and filter such material. The reason for this is given by the fact that such content is often prohibited in certain environments (e.g., schools and workplaces) or certain publics (e.g., children). In recent years, many works of the literature have been mainly focused on detecting pornographic images and videos based on visual content, particularly on the detection of skin color. Although these approaches provide good results, they generally have the disadvantage of a high false positive rate, since not all images with large areas of skin exposure are necessarily pornographic images, such as people using swimsuits or images related to sports. Local features based approaches, with Bag-of-Words models (BoW), have been successfully applied to visual recognition tasks in the context of pornography detection. Despite existing methods provide promising results in the context of detection of pornography, they use local features descriptors that require a high computational processing time, and generate high-dimensional vectors. In this work, we propose a simple, effective and efficient approach to the problem of visual recognition in the context of pornography detection. The method is based on local features extraction using binary descriptors, a low-complexity alternative, in conjunction with the recent mid-level representation BossaNova, a BoW model extension that preserves a more richly visual information. The results validated the proposed approach by presenting results with superior quality compared to other approaches in the literature.