Research On Technology Of Network Image Content Filtering

With the rapid development of the internet and netizens, internet as a new way for people to obtain a lot of information, plays a huge role in broadening their horizons, foreign exchange and promoting youth personality development. All kinds of sensitive information more and more by means of text, pictures, video to spread, brings people especially young people great negative impact. The detection and filtering for internet information have an urgent need to develop effective tools.In order to prevent the spread of sensitive images effectively, this paper presents a method for content filtering of the sensitive image. The approach establishes an RGB, YUV and YIQ mixed model to detect the exposed areas of skin. To reduce the misjudged probability of a normal image effectively, we need to extract the maximum connected region and detect the face.The important feature of sensitive image contains large area of skin exposed, and accurately detecting of the skin exposed areas lays a good foundation for later analysis and identification. Mixed RGB, YIQ, YUV color model can be well found region. By mixed model to detect skin, we can obtain a binarized image mask. Based on the image mask to extract features, that including skin area percentage, the percentage of region area, relevant characteristics of connected region and so on, which are conducive to filter sensitive image.The statistical analysis for sensitive images finds that the majority of sensitive images contain face. That is an important feature for sensitive image and you can also use it to distinguish scenery image from skin color area. By analyzing the technology of skin model of sensitive image, mask image and face detection, we implement a content detection of sensitive image system. Experimental results show that this paper proposes a filtering technology of network image content can better complete filtering of sensitive images, and have more than80%accuracy.