Title :
The Mask-SIFT cascading classifier for pornography detection
Author_Institution :
Dept. of Comput. Sci., Virginia Polytech. Inst. & State Univ., Falls Church, VA, USA
Abstract :
Pornography detection using the Scale Invariant Feature Transform (SIFT) has been shown effective in identifying pornographic images. By including automated Gaussian skin masking for feature isolation, classifier performance is significantly improved. Similarly, utilizing a cascading classifier that pre-filters images based on size and skin percentage further improves precision and recall with a substantial increase in classification speed.
Keywords :
Gaussian processes; feature extraction; image classification; object detection; automated Gaussian skin masking; feature isolation; mask-SIFT cascading classifier; pornographic image identification; pornography detection; scale invariant feature transform; Classification algorithms; Detectors; Feature extraction; Image color analysis; Internet; Skin; Training; Pornography detection; feature recognition; skin detection;
Conference_Titel :
Internet Security (WorldCIS), 2012 World Congress on
Conference_Location :
Guelph, ON
Print_ISBN :
978-1-4673-1108-3
