A NOVEL MULTIMEDIA DATABASE SYSTEM FOR EFFICIENT IMAGE/VIDEO RETRIEVAL BASED ON HYBRID-TREE STRUCTURE
With recent advances in computer vision, image processing and analysis, a retrieval process based on visual content has became a key component in achieving high efficiency image query for large multimedia databases. In this paper, we propose and develop a novel indexing mechanism built on the top of a Xindice database targeted at supporting quick access to image/video data along with strong retrieval functionalities for those high-dimensional metadata. The proposed database applies a native XML database as a backbone, uses hybrid-tree as its database structure, and performs Threshold Algorithm (TA) to conduct similarity search based on a random combination of image features. The database also sketches an index scheme in building a mapping between temporal and spatial domains so as to support both temporal and spatial queries at a higher speed search over the traditional pure temporal or spatial based indexing methods.The simulation results demonstrate that the proposed database system can conduct a more efficiently search activities supporting queries that involve composite features with arbitrary Boolean combinations and can outperform a traditional database employing a linear scanning search scheme.
Multimedia database hybrid-tree PCA temporal & spatial indexing
JING CHEN JIA-LIE SHEN JIAN ZHANG KIANDY WANGSA
National ICT Australia, 223 Anzac Parade, Kensington NSW 2052, Australia
国际会议
2006 International Conference on Machine Learning and Cybernetics(IEEE第五届机器学习与控制论坛)
大连
英文
4353-4358
2006-08-13（万方平台首次上网日期，不代表论文的发表时间）