Internet Supported Analysis of MPEG Compressed Newsfeeds

In this paper new ideas and algorithms for adapting the workflow for newscast production are presented. The goal is to enable TV news editors to take advantage of the added facilities offered by digital video techniques. Algorithms for the analysis of MPEG compressed videos are presented. These enable a news editor to extract sequences of key frames from newsfeeds. Newsfeeds consist of assembled news clips transmitted by news agencies. Key frames are extracted in order to create a quick overview of the available material. In addition frames containing textual descriptions of the news content of individual clips are also extracted. These images are of low quality and standard OCR methods prove unsuitable for text recognition. Using standard image processing techniques in combination with a new OCR method especially developed for this purpose key words are extracted from these frames. On the one hand the key words can be used by news editors to search for particular news items in a newsfeed and on the other to search for related information on the internet. The final output of the system thus supplies news editors with an overview of newsfeed content combined with additional information gathered from news agencies via the internet.