Nude image classification using rough set analysis

Nude image distributions on Internet become a serious problem for many children and their parents around the world. In this paper, a rough set analysis technique is applied to deal with vagueness information of nude image. Many research papers have been published how to detect a nude image using skin color, texture or shape measures. The pixel features including YCbCr range, area of human skin, eccentricity, are chosen as nude image features because of their easy acquisition. The YCbCr range is Y > 100, 100 <; Cb <; 120 or 128 <; Cb <; 130 and 130 <; Cr <; 160. The purpose of this paper is to classify a certain or probable nude image in a simple if-then rule. The accuracy of experimental result is 95% with a rough set analysis and 85.2% without a rough set analysis.