Feature extraction methods for cartoon character recognition

This paper describes feature extraction methods for an eye and hair to recognize cartoon characters. Image data sharing sites, such as YouTube, have numerous pictures that infringe on copyrights. To solve this problem, we study a method that recognizes authors or characters in cartoons and classifies the pictures infringing copyright. This paper proposes feature extraction methods for an eye and hair for cartoon character recognition. Experimental results show that black pixel density and peripheral direction contributivity features of an eye and stroke density features of hair are effective for cartoon character recognition.