Two-dimensional entropic segmentation

A novel method of 2-D entropic segmentation using a linear discriminant function is presented. This segmentation method automatically highlights desired objects against background with no user input or parameter specification. Improved class separation, and therefore, reduced classification error can be obtained by adding a second feature to the gray level histogram. Examples of aerial and medical images were segmented using features such as intensity, and fractal error or standard deviation with lower segmentation errors than 1-D and 2-D entropy-based methods. (C) 1999 Elsevier Science B.V. All rights reserved.