Performance evaluation of text detection and tracking in video

摘要: Text detection and tracking is an important step in a video content analysis system as it brings semantic clues which vital supplemental source of index information. While there has been significant amount research done on text tracking, are very few works performance evaluation such systems. Evaluations this nature have not attempted because the extensive effort required to establish reliable ground truth even for moderate dataset. However, ventures gaining importance now. In paper, we propose generic method object systems domains where objects can be bounded by simple geometric shapes (polygons, ellipses). Two comprehensive measures, one each proposed substantiated capture different aspects task single score. We choose tasks show effectiveness our framework. Results presented from evaluations existing algorithms using real world data metrics shown effective measuring total accuracy these algorithms.