Search operations in databases require special support at the physical level. This is true for conventional databases as well as spatial databases, where typical search operations include the point query (find all objects that contain a given search point) and the region query (find all objects that overlap a given search region). More

We propose a method to achieve declustering for cartesian product files on M units. The focus is on range queries, as opposed to partial match queries that older declustering methods have examined. Our method uses a distance-preserving mapping, namely, the Hilbert curve, to impose a linear ordering on the multidimensional points (buckets); then, it traverses the buckets according to this ordering, assigning buckets to disks in a round-robin fashion. Thanks to the good distance-preserving properties of the Hilbert curve, the end result is that each disk contains buckets that are far away in the linear ordering, and, most probably, far away in the k-d address space. This is exactly the goal of declustering. Experiments show that these intuitive arguments lead indeed to good performance: the proposed method performs at least as well or better than older declustering schemes. Categories and Subject Descriptors: E.1 [Data Structures]; E.5 [Files]; H.2.2 [Data Base Management]: Physical Des...

In this paper we present a new technique to improve the performance of spatial access methods by minimizing redundancy: the oversize shelf . Oversize shelves are additional disk pages that are attached to the interior nodes of a tree--based spatial access method (such as the R + --tree or the cell tree). These pages are used to accommodate very large data objects in order to avoid their excessive fragmentation. Whenever inserting a new object into the tree, one now has to decide whether to store it on an oversize shelf or insert it into the corresponding subtrees. For this purpose, we developed an analytic model for the behavior of dynamic spatial access methods under insertions and deletions. The model yields a threshold value for the size of an object, such that it is more favorable to put it on the oversize shelf if and only if its size is greater than the threshold value. Otherwise the insertion into the corresponding subtrees is preferable. Practical experiments indicate that th...

Abstract not found

This thesis describes research into an area of content based image retrieval (CBIR), that of feature indexing for the purpose of rapid retrieval. The techniques in this thesis draw from the field of text IR and demonstrate that individual image extraction algorithms can be optimised for use with an inverted index, which could lead to CBIR systems capable of sub-second retrieval times on collections of millions of images. A novel global feature algorithm...