Sensitivity of Area–Perimeter Relation for image analysis and image segmentation purposes

Abstract

Image analysis with the use of fractal estimators is important for the description of grayscale images. The sensitivity of Area–Perimeter Relation (APR) using Brodatz texture database and Monte Carlo approach is evaluated in this paper. Obtained APR curve is approximated using polynomial and two parameters of polynomial are applied as discrimination parameters. A few techniques for the evaluation of APR are applied. The results show the possibility of the discrimination using single or two polynomial parameters even for a few textures. The quality of discrimination (separation between textures classes) could be improved if larger window analysis sizes is applied.