Share Email Print

Proceedings Paper

Sensitive analysis and segmentation of low-contrast images: knowledge discovery based on multiparameter topological resonance method
Author(s): Alexander M. Akhmetshin; Lyudmila G. Akhmetshina
Format Member Price Non-Member Price
PDF $14.40 $18.00
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

A new method of low contrast images analysis and segmentation is outlined. The one has providers high sensitivity to detection of visually invisible low contrast areas and simultaneous stability to influence of local small brightness variations. The new method consists of the following four steps. 1) Forming a moving window of size (3x3). Analyzed image brightness pixels into the bounds of the window are considered as coefficients some virtual nonrecursive digital filter. The one is characterized by its spectral characteristic. It gives possibility for comparing to each pixel of analyzed image its virtual spectral characteristic. 2) An information features for futher analysis there are using resonance points (magnitudes and frequencies) of virtual magnitude-frequency and group-delay characteristics (namely it stipulated using the method designation as multiparameter topological resonance one). 3) Visualization new synthesized features and a separate analysis of the new images. 4) Multiparameter information fusion into one resulting image (if it is needed) on base self-organizing map method.On base our experimental investigations it had been established importance virtual group-delay function (i.e. new information characteristic) for tasks detection and segmentation low contrast fuzzy regions of analyzed images. Experiments were made with examples of real low contrast images: X-ray CT, digital mammograms, and geophysical field images. For all situations there were obtained very good practical results.

Paper Details

Date Published: 12 April 2004
PDF: 8 pages
Proc. SPIE 5433, Data Mining and Knowledge Discovery: Theory, Tools, and Technology VI, (12 April 2004); doi: 10.1117/12.540904
Show Author Affiliations
Alexander M. Akhmetshin, Dniepropetrovsk National Univ. (Ukraine)
Lyudmila G. Akhmetshina, Dniepropetrovsk National Univ. (Ukraine)

Published in SPIE Proceedings Vol. 5433:
Data Mining and Knowledge Discovery: Theory, Tools, and Technology VI
Belur V. Dasarathy, Editor(s)

© SPIE. Terms of Use
Back to Top