The «TerraRush» project has been selected together with 18 others (including the project « eTourism Augmented Reality » you have read about in our February Newsletter) among over 70 applications. This is the result of the first call for proposals «Scanning technologies and enhancement of cultural, scientific and educational content » within the framework of the Investment Program for the Future launched by the French government.
« TerraRush » aims to produce a platform with technological solutions able to:
- execute complex searches in an important volume of video data ;
- carry out editing tasks as well as selective and automated transcoding in order to produce new contents that are exploitable for broadcasting, with adaptive formats according to the network capacity and the constantly changing multimedia terminals (from HD to the simplest smart phone videos) ;
- integrate innovations on the web in terms of new distributed services and applications.
This platform will allow us to industrialize the process of indexing video files from digitized rushes (original audio and video documents, that are raw and unedited) thanks to a semi-automated process that facilitates to place new services on the market:
- sale of extracted program sequences;
- sale of stock footage;
Without this platform many of those rushes (over 150 000 hours) will be destroyed for lack of economic solutions allowing using them. Their holders want to make room for their new HD digital productions. Nevertheless the rushes, witnesses of the years 1970 – 2010, represent a rich heritage and great potential.
The new platform will include implementations of sophisticated innovations such as:
- multimedia search techniques by content (namely based on object recognition and automated classification of sound characteristics and visual features of audio and video data stream);
- techniques for massive data processing on parallel and distributed architectures;
- development of innovative Web services intended for content management and broadcasting.
Within this project the LRDE will carry out text extraction in image sequences in order to enrich the databases of rushes. The automatic extraction of text helps indexing the rushes. In the same time, the LRDE will work on removing text from the sequences allowing restoring the original image.
This functionality is useful when you want to erase the subtitles for instance, and recover the video sequence.