Pornographic Image Detection Based on Multilevel Representation

Yushi Wang,Qingming Huang,Wen Gao
DOI: https://doi.org/10.1142/s0218001409007739
IF: 1.261
2009-01-01
International Journal of Pattern Recognition and Artificial Intelligence
Abstract:With the proliferation of pornographic images on the Internet, it is essential to automatically detect pornographic images by analyzing image content. Most traditional detection systems are based on low-level features and generate many false positives due to images that contain large regions of skin-like colors. In this paper, we present a novel detection method based on local features, such as SIFT (Scale Invariant Feature Transform) visual words. Support Vector Machine (SVM) is used to classify images according to their multilevel representation based on visual words and the distribution of pornography-related visual words. The multilevel representation captures inter-word statistics and fuses various visual components of pornographic scenes. Experimental results demonstrate that our method outperforms traditional skin-region and human-body-model based methods, and performs well on a wide range of test data, in particular, on human-related images.
What problem does this paper attempt to address?