Pornographic web content filtering with skin color information

In this study, a system has been designed and implemented for filtering pornographic web sites automatically. Usually, text information is used for web content filtering. However, web pages especially health pages can also be filtered by text information filtering technique. Hence, studies about web page filtering via image information is in progress nowadays. Detection of skin color pixels from images is a quite difficult problem because of the conditions that images taken from but skin detection technique is frequently used for detecting nudity images. Color spaces are important because of the performance. In this study, RGB, HSV and RGB & HSV color spaces have been selected. Artificial Neural Networks were used to classify the nudity pictures and a software named GezgIN was developed. Experimental results have shown that, RGB & HSV color spaces results are more effective than the results of the usage of these color spaces individually, GezgIN has generated appropriate results as long as image count on a web site is lower than 50 and each image size on this site is about 200x200.