This paper presents a new method for estimating the fractal dimension of on
e-dimensional profiles. In this approach, the real fractal dimension D is c
onsidered as an implicit continuous function of the estimated fractal dimen
sion D-e, the resolution and several other elements. By approximating this
function from a number of experimental data, we can obtain more precise est
imates of the fractal dimension D. This approximation is done using a fuzzy
logic controller and an averaging procedure, permitting to, respectively,
decrease two kinds of estimation errors: (1) systematic errors, which are a
ssociated with values of D, resolution, trends of profiles, and etc. (2) st
ochastic errors, which are mainly caused by the choice of the sequence {eps
ilon (k)} representing the sizes of structuring elements corresponding to d
ifferent scales. The effectiveness of this method is shown by estimating fr
actal dimensions for two sample functions and a number of natural and synth
etic fibers. (C) 2000 Pattern Recognition Society. Published by Elsevier Sc
ience Ltd. All rights reserved.