Balancing Strategies For Parallel Content-Based Data Retrieval Algorithms In A K-Tree Structured Database

Radu Dobrescu, Matei Dobrescu, Daniela Hossu
2010 Zenodo  
The paper proposes a unified model for multimedia data retrieval which includes data representatives, content representatives, index structure, and search algorithms. The multimedia data are defined as k-dimensional signals indexed in a multidimensional k-tree structure. The benefits of using the k-tree unified model were demonstrated by running the data retrieval application on a six networked nodes test bed cluster. The tests were performed with two retrieval algorithms, one that allows
more » ... e that allows parallel searching using a single feature, the second that performs a weighted cascade search for multiple features querying. The experiments show a significant reduction of retrieval time while maintaining the quality of results.
doi:10.5281/zenodo.1076283 fatcat:fpx5t7r4xfdivnptmddge2z5ce