14 March 2013 An effective method on pornographic images realtime recognition
Author Affiliations +
Proceedings Volume 8768, International Conference on Graphic and Image Processing (ICGIP 2012); 876862 (2013) https://doi.org/10.1117/12.2006020
Event: 2012 International Conference on Graphic and Image Processing, 2012, Singapore, Singapore
Abstract
In this paper, skin detection, texture filtering and face detection are used to extract feature on an image library, training them with the decision tree arithmetic to create some rules as a decision tree classifier to distinguish an unknown image. Experiment based on more than twenty thousand images, the precision rate can get 76.21% when testing on 13025 pornographic images and elapsed time is less than 0.2s. This experiment shows it has a good popularity. Among the steps mentioned above, proposing a new skin detection model which called irregular polygon region skin detection model based on YCbCr color space. This skin detection model can lower the false detection rate on skin detection. A new method called sequence region labeling on binary connected area can calculate features on connected area, it is faster and needs less memory than other recursive methods.
© (2013) COPYRIGHT Society of Photo-Optical Instrumentation Engineers (SPIE). Downloading of the abstract is permitted for personal use only.
Baosong Wang, Xueqiang Lv, Tao Wang, Chengrui Wang, "An effective method on pornographic images realtime recognition", Proc. SPIE 8768, International Conference on Graphic and Image Processing (ICGIP 2012), 876862 (14 March 2013); doi: 10.1117/12.2006020; https://doi.org/10.1117/12.2006020
PROCEEDINGS
5 PAGES


SHARE
Back to Top