A Novel Method for Movie Character Identification Based on Graph Matching: A Survey

Abstract

Automatic face identification of character in movies received tremendous attention from both video content understanding and video annotation because of their application in movie industry such as video semantic analysis, video summarization, and personalized video retrieval. Character identification of movie is challenging problem due to huge variation in the appearance of each character and complex background, large motion, non-rigid deformation, occlusion, huge pose, expression, wearing, clothing, even makeup and hairstyle changes and other uncontrolled condition make the result of face detection and face tracking unreliable. In particular, character identification for movie used video and script. Face tracking and clustering from video and name of person extract from script. Many challenges for face clustering and face-name matching are present. In good situation and clean environment existing methods gives better result, but in a complex movie scene performance is limited because face tracking and clustering process generate a noise. In this paper we present a comparative study of three methods using textual cues like cast list, script, subtitle and closed caption based on local and global face-name matching