Spatiotemporal restoration of regions occluded by text in video sequences

Occasionally, there is a case in which text inserted manually in a video sequence for convenience of viewers is unnecessary for reuse of that video. For reusability, we propose an approach for automatic detection and removal of caption text using an SVM and spatiotemporal restoration in video sequences. After detecting text using an SVM, we restore the regions occluded by the text using a spatiotemporal restoration scheme. The proposed approach includes many applications such as translation of captions and replacement of indirect advertisements in videos.