content based video/image indexing and retrieval
We have developed – in cooperation with the Image Processing Laboratory of the Dept. of Image Processing and Neurocomputing at the University of Pannonia, Hungary - multiple image and video feature extraction methods which can be used to search for similar images and videos in databases, and a search&retrieval system for performing such searches. First videos are imported in a database backend, shots and representative frames are extracted, optional textual annotations are added, then content features are automatically extracted without user interaction.
After the feature extraction an index of the database contents is updated and searches can be performed. Queries can be formulated with text and by providing a sample image. Currently we support around a dozen features which can be used during the similarity search. (Re)Indexing can be a timely process, while searching is almost real-time.
The system’s main parts are
- a server side backend for importing videos, registering feature extractor modules, managing annotations
- az indexer application
- a query service which performs the content-based searches
- a text-based search module
- a web-based user interface for formulating the queries and presenting the results
Some shots of the system in action are shown below.
A screenshot demo video showing search in action can be downloaded HERE (you will need the xvid codec, or ffdshow decoder to view).
Results of a textual search.
When clicking on a result, the shot's details, annotations and the video part itself can be viewed.
Results of a content based query (using edge based features).
Results of a content based query.Contact information