Filmmakers can identify faults during production with an automatic fault identification and quality assessment system for video sequences. Scientists are also presenting automatic classification for videos at the International Broadcasting Convention in Amsterdam.
Some typical faults that the software of the Fraunhofer Institute for Digital Media Technology IDMT identifies are outputting mono instead of stereo sound, getting channels confused or undesired fluctuations in volume. In addition, it can find a blur, freezes or over/underexposure. At present, developers are working at identifying grainy pictures, visible coding artifacts, a faulty 16:9 / 4:3 switch-over, color distortion or problems with synchronity. Automatic fault and quality monitoring means enormous savings in time especially for production companies when viewing video material because they can automate steps in work and save resources.
Beyond this, the experts of IDMT will also be presenting a technology for automatically classifying video material in semantic categories at the IBC. This software analyses videos for certain characteristics such as genre or mood. That makes it possible for TV viewers to find videos in the thriller category in a large range of films. This technology can be used as a recommendation system for films in mediatheques and video-on-demand portals and broadcasters can also streamline their work routines - classified video material can be found more quickly and archived easier.
The special thing about this analysis and recommendation system is the fact that it works fully automatically. Instead of manually entering key words, a semantic analysis allows an automatic description of the genre and mood while new search functionalities make it possible to put together your personalized entertainment program.
The IDMT researchers will be showing how software identifies faults in video sequences, assesses quality and automatically classifies videos at IBC (Hall 8, Stand C81) from September 10-14.
Explore further: Managing the Internet of Things