ABSTRACT
The theory of moments [8] provides a convenient way of characterizing patterns within images. We are interested in using them as descriptors for each
1 INTRODUCTION
Spatial domain filters are the most direct way to characterize edges, granularity and shape as global features within image analysis. An important subset of spatial domain filters are based on moment functions of pixel intensity values [9]. These functions are expressed as discrete sums, with some of them being invariant under image translation, scale change and rotation [2,4]. Moments with orthogonal basis functions such as Legendre and Zernike polynomials are even more useful as image descriptors because they can represent the image by a set of mutually independent descriptors, with a minimal amount of information redundancy [8].
