A Relaxed Algorithm for Similarity Queries Performed with High-Dimensional Access Structures
Henrich, Andreas (2025): A Relaxed Algorithm for Similarity Queries Performed with High-Dimensional Access Structures, in: Bamberg: Otto-Friedrich-Universität, S. 376–390.
Details
- Faculty/Chair
- Faculty/Professorship
- Author
- Publisher Information
- Otto-Friedrich-Universität : Bamberg
- Year of publication
- 2025
- Source/Other editions
- Akmal B. Chaudhri, Rainer Unland, Chabane Djeraba, u. a. (Hrsg.), XML-Based Data Management and Multimedia Engineering : EDBT 2002 Workshops ; revised papers, Berlin u.a.: Springer, 2002, S. 376–390, ISBN: 978-3-540-00130-0
- Year of first publication
- 2002
- Language
- English
- Licence
- URN
- Abstract
Similarity queries searching for the most similar objects in a database compared to a given sample object are an important requirement for multimedia databases. However, strict mathematical correctness is not essential in many applications of similarity queries. For example, if we are concerned with image retrieval based on color and texture similarity, slight mathematical inaccuracies will hardly be recognized by the human observer. Therefore we present a relaxed algorithm to perform similarity queries for multidimensional index structures. This algorithm assures only that a user defined portion of the result list containing n elements actually belongs to the n most similar objects — the remaining elements are subject to a best effort semantics. As we will demonstrate, this allows to improve the performance of similarity queries by about 25 % with only marginal inaccuracies in the result.
- GND Keywords
- AlgorithmusÄhnlichkeitssucheMehrdimensionale Datenstruktur
- Keywords
- Relaxed Algorithm
- DDC Classification
- 004 Computer science
- RVK Classification
- ST 270
- Type
- Conferenceobject
- URI
- Activation date
- 2025-06-02