A Test-bed Facility for Research in Video Database Benchmarking

Purdue University, West Lafayette IN

Investigators

Abstract

This project will create a comprehensive test-bed facility to support research in video database management. Real-world video applications demand database technology that is capable of providing real-time content-based query and retrieval for stored digital video. However, video is a complex data type that requires a significant amount of preprocessing for content representation before research issues in database management for real-time video data can be addressed. This project will facilitate video database research by abstracting video as a fundamental and fully functional database data type. Tools for video content analysis, feature extraction and high-dimensional indexing will be integrated into the test-bed facility, which will then be used to address several research challenges in the management of real-time streamed video data, including multi-feature nearest neighbor queries, feature-based query processing, video operators, buffer management for streaming video, and continuous video streaming with real-time constraints. The system will provide standardized methods for integrating components and measuring their performance, and will be furnished with a large collection of indexed videos. The underlying database management system, video-processing tools, and video data will form a complete environment for the video database research community, providing an effective platform for implementation, validation and evaluation of new video database technology.