Performance of SS-Tree with Slim-Down and Reinsertion Algorithm

Along with the development of information technology, plenty of multimedia data appears. The growth of these data brings the need for more effective methods in retrieval. Multimedia retrieval systems always index these data based on feature vectors. And the index structures such as the R-tree family, are used to manage these feature vectors more efficiently. Slim-down algorithm is used in Slim-tree, and it can improve the number of disk accesses for range queries in average 10%-20% for vector datasets. Especially for datasets with bigger bloat-factors, the average improvement goes to 25%-35%. In this paper, we use slim-down algorithm in SS-tree index structure and compare it with reinsertion algorithm, to test whether it´s efficient for improving the performance of SS-tree and outperforms Reinsertion algorithm. Experiment results show that Slim-down algorithm can effectively reduce the overlapping region between nodes, thus the performance of SS-tree gets improvement. And for the feature vectors with high dimension, it outperforms Reinsertion algorithm.