Entity Detection for Information Retrieval in Video Streams
The growing amount of video data has raised the need for automatic semantic information indexing and retrieval systems. To accomplish to these needs, the text information in images and videos is proved to be an important source of high-level semantics. This paper discusses the video OCR system designed for overlay text based automatic indexing and retrieval in the video streams. The proposed framework consists of the video segmentation, the video key-frame extraction, the video text recognition, and the entity detection. The experimental results on Korean television news programs show that the proposed method efficiently realizes the automatic indexing in the video streams.
KeywordsOverlay text Video OCR Indexing Named entity NLP
This research was supported by the MSIT (Ministry of Science and ICT), Korea, under the Grand Information Technology Research Center support program (IITP-2018-2016-0-00318) supervised by the IITP (Institute for Information & communications Technology Promotion).