Share Email Print

Proceedings Paper

Content-based retrieval of distorted images using a hybrid genetic algorithm augmented by a self-organizing network
Author(s): Igor V. Maslov
Format Member Price Non-Member Price
PDF $17.00 $21.00
cover GOOD NEWS! Your organization subscribes to the SPIE Digital Library. You may be able to download this paper for free. Check Access

Paper Abstract

Content-based image retrieval involves a direct matching operation between a query image and a database of stored images. In case where the query image can be significantly distorted in relation to the stored image, the common methods of computing similarities between the feature vectors of the images might not provide sufficiently robust base for successful image retrieval. The retrieval problem can be re-formulated then as an optimization problem of finding a correct mapping between images when a search space is considerably large. An approach is proposed that combines a hybrid genetic algorithm and a self-organizing neural network in a global search for a set of parameters defining a correct mapping between images. In order to compute unique characteristics of the query image that are invariant to image transformation and distortion, an image transform is introduced in the form of image local response. The transform is applied at the pre-processing stage to the stored and the query images, to extract their dynamic content. Correlation is applied then to both images in the response space. The technique reduces the search space and allows to find a set of subregions that can contain a potentially correct match. After pre-processing, a hybrid genetic algorithm augmented by a self-organizing network for local refining finishes the search in the reduced parameter space.

Paper Details

Date Published: 26 November 2003
PDF: 12 pages
Proc. SPIE 5242, Internet Multimedia Management Systems IV, (26 November 2003); doi: 10.1117/12.510797
Show Author Affiliations
Igor V. Maslov, CUNY/Graduate Ctr. (United States)

Published in SPIE Proceedings Vol. 5242:
Internet Multimedia Management Systems IV
John R. Smith; Sethuraman Panchanathan; Tong Zhang, Editor(s)

© SPIE. Terms of Use
Back to Top