A novel classificatory filter for multiple types of erotic images based on Internet

In the condition of which various types of erotic images are available on internet, so big an image database make the retrieval precision and recall rate stay low. Considering the fact that the present erotic image filtering algorithms has a higher false-positive rate when the image is of dark brightness and color distortion, a kind of classificatory filter for multiple types of erotic images on Internet was proposed. The filter consists of three parts: skins-color model matching layer, dark erotic image detecting layer and a decision-making layer based on SVM. The experiment results show that the model can simply realization, overcome the shortness of the traditional methods, effectively improve the detection correction, and fit the requirements of application on Internet.