An Adult Image Recognition Method Facing Practical Application

Abstract

In this paper, for practical application about adult image recognition, features of the false positive images are analyzed, a new adult image recognition method is presented facing practical application. Firstly obvious normal images are filtered out making use of skin detection based on wavelet texture, secondly face analysis and color consistent image analysis are used to filter normal images satisfying the rule, finally the recognition result is obtained making use of false positive images comparison technology based on E2LSH. The experimental results show that this method can reduce false-positive rate greatly on the premises that remain the recognition rate fairly constant, compared with traditional methods.