An algorithm of pornographic image detection

Authors:
Hong Zhu;Shuming Zhou;Jianying Wang;Zhongke Yin
Affiliations:
Chinese Academy of Sciences/The Chinese University of Hong Kong, China;Chinese Academy of Sciences/The Chinese University of Hong Kong, China;Southwest Jiaotong University, China;Southwest Jiaotong University, China
Venue:
ICIG '07 Proceedings of the Fourth International Conference on Image and Graphics
Year:
2007

Citing 0
Cited 4

Malicious content filtering based on semantic features

Proceedings of the 2nd International Conference on Interaction Sciences: Information Technology, Culture and Human
Real-time detection of children's skin on social networking sites using Markov random field modelling

Information Security Tech. Report
Automatic nipple detection using shape and statistical skin color information

MMM'10 Proceedings of the 16th international conference on Advances in Multimedia Modeling
On the multi-agent learning neural and Bayesian methods in skin detector and pornography classifier: An automated anti-pornography system

Neurocomputing

Quantified Score

Hi-index	0.00

Visualization

Abstract

Real-time detection of pornographic images can effectively prevent pornographic images from spreading on the Internet. The methods of detection mainly focus on the identification of skin region. To overcome the disadvantages of traditional algorithms, in this paper, we propose a skin model based on the combination of YIQ, YUV, and HSV. In the step of the pre-dealing, we use white balance algorithm to achieve better skin area. Then, texture model based on Gray Level Co-Matrix (GLCM) and geometric structure of human beings are used to decrease the disruptions of the background region similar with the skin area. The features which extracted from the last images dealt by color and texture model are input into Support Vector Machines (SVM), through which the pornographic images and assorted control images are classified successfully. An experiment using 400 pornographic images and 400 assorted control images is carried out. Experimental results show that the algorithm of skin texture-based and feather-based detection has perfect performances, and the detection rate can reach 88.89%.