Research And Applications Of Key Techniques On Image Filter

How to keep eroticism off Internet is an important technique problem. It is not sufficient to deal with this problem by blockading Web addresses and filtering erotic texts. In order to prevent the spread of eroticism effectively, it is necessary to analyze the images embed in the homepage by the computer vision technique. Founded on "the key technology research of image searching and filtering in real time" of 863-plan project, we study the key technology of Image Filter. In this paper, we present a new method of filtering objectionable images and develop a system which can filter the erotic images efficiently. The results show that the system works well.In this paper, all the work we have done is based on the image bank which we have constructed. Among the work, the most important one is the Skin Color Model by which we get skin color masks from original images. Nevertheless, it's not very precise to locate the skin color region only in this way. So the Texture Model is introduced into the system. Consequently, we get the skin region by the combination of Skin Color Model and Texture Color Model. Based on the work above, we extract the features of Skin Color Mask Images and Skin Mask Images to classify the images by classifier.At the stage of skin color processing, we develop a new skin color model, Statistical Chroma Space Model(SCSM). In addition, we present the methods of how to evaluate the separability of image bank and how to estimate GMMparameters rapidly. Our work at the stage of texture processing is to choose an appropriate Texture Model by studying the former work. Our work at the stage of classifier processing is building feature vectors and proving its effectivity by experiments, choosing the appropriate classifier by studying the former work.