text detection and recognition in arabic news video

Recognizing texts in video, often called VIDEO-OCR represents an active research domain. As an application of such field we cite video classification, content-based video retrieval, automatic broadcast annotation,etc. The end goal of my thesis aims to provide an automatic solution for broadcast annotation. To do this, we need to develop a standard dataset to train and evaluate the existing video-ocr systems, an automatic video text detector and an OCR system to automatically recognize the already detected text regions. Our potential clients are the archivist and documentalist of news TV channels.