A pornographic image filtering model based on erotic part

Compared with some deficiencies of the traditional filtering models using the features of skin color only, this paper presents a pornographic image filtering model based on erotic parts detection. And the information of face, skin color and trunk contour is utilized to implement the erotic part recognition. A decision tree classifier is also employed to identify porno image based on skin color and erotic part features. The extensive experiments indicate that this porno image detection model provides a higher detection rate and a lower error rate than previous algorithms. Its correct detection rate reaches 89.50% in the detection of 505 porno images and its error rate is 18.64% in the detection of 971 benign images with bare person.