Trail-based approach for video data indexing and retrieval

Increasing use of multimedia data makes it crucial to use intelligent search mechanisms for retrieving multimedia data by content. Digital video requires the incorporation of temporal information for any effective content-based retrieval scheme. We present a trail-based model which uses the object motion information in order to characterize the events for subsequent searches for "similar" clips. Algorithms for different spatio-temporal search modes using various digital signal processing techniques are also introduced. We implemented the proposed methods and demonstrated that high-level query formulation can be achieved for the aforementioned purpose.