Texture mapping in image segmentation

Document Type

Conference Proceeding

Date of Original Version

12-1-1991

Abstract

The use of fractals in image analysis is examined in the context of segmenting cardiac magnetic resonance (MR) images via the mapping of textures. These images are assumed to be composed of multiclass fractional Brownian motion (FBM), and from this assumption a parameter H, the Hurst coefficient, is extracted by the use of a maximum likelihood estimator (MLE). The MLE is applied to a probability density function (PDF), which is dependent upon the distribution of the noise within small regions of the image. The results, as applied to patient data, are also discussed, in particular the complementary information yielded when this method of segmentation is compared with the segmentations produced by gradient and statistical Gaussian operations.

Publication Title, e.g., Journal

Proceedings of the Annual Conference on Engineering in Medicine and Biology

Volume

13

Issue

pt 5

This document is currently not available here.

Share

COinS