Fourier theory provides an important approach to shape analyses; many
methods for the analysis and synthesis of shapes use a description bas
ed on the expansion of a curve in Fourier series. Most of these method
s have centered on modeling regular shapes, although irregular shapes
defined by fractal functions have also been considered by using spectr
al synthesis. In this paper we propose a novel representation of irreg
ular shapes based on Fourier analysis. We formulate a parametric descr
iption of irregular curves by using a geometric composition defined vi
a Fourier expansion. This description allows us to model a wide variet
y of fractals which include not only fractal functions, but also fract
als belonging to other families. The coefficients of the Fourier expan
sion can be parametrized in time in order to produce sequences of frac
tals useful for modeling chaotic dynamics. The aim of the novel charac
terization is to extend the potential of shape analyses based on Fouri
er theory by including a definition of irregular curves. The major adv
antage of this new approach is that it provides a way of studying geom
etric aspects useful for shape identification and extraction, such as
symmetry and similarity as well as invariant features.