Rough Set Based Objectionable Image Recognition

Gy Zhu,Hx Yao
2003-01-01
Abstract:In this paper we introduce rough set theory into image object classification and apply this idea to solve the problem of objectionable image recognition. We propose a new method for objectionable image recognition corresponding to the image object content based on rough set. The method consists of an image feature extractor based on the segmented skin region objects passed by the skin detection and a trained classifier utilizing the idea of rough set theory. Our system has achieved 90.9% recognition rate for objectionable images with 3.5% false alarm rate. It is demonstrated that the proposed method is effective and efficient. The framework of our proposed method can be easily extended to apply for other recognition applications based on special object classification utilizing rough set theory.
What problem does this paper attempt to address?