Abstract

In image database systems, a perfect hash table can be used to enhance the efficiency and effectiveness of the image retrieval process. In our earlier work, we have proposed heuristic algorithms to compute the minimal perfect hash table from symbolic picture objects. The hash table thus computed cannot be modified easily, limiting its application to fixed databases like the ones on CD-ROMs. In this paper, we propose a new data structure to allow for insertion and deletion in the hash table. The new data structure, the near-perfect hash table, allows a limited number of collisions on some of the hash addresses. The near-perfect hash table provides the flexibility of a live database while keeping intact the retrieval efficiency of the perfect hash table.

Department(s)

Computer Science

Keywords and Phrases

2D string; Hash table; Image database indexing

Document Type

Article - Conference proceedings

Document Version

Citation

File Type

text

Language(s)

English

Rights

© 2024 Association for Computing Machinery, All rights reserved.

Publication Date

18 Feb 1996

Share

 
COinS