In this paper, a novel texture classification scheme using higher-orde
r statistics (HOS) functions as discriminating features is proposed. I
t is well known that such statistical parameters are insensitive to ad
ditive Gaussian noise. In particular, third-order statistical paramete
rs, i.e. third-order cumulants and bispectrum, are insensitive to any
symmetrically distributed noise, and also exhibit the capability of be
tter characterizing non-Gaussian signals. By exploiting these HOS prop
erties, it is possible to devise a robust method for classifying textu
res affected by noise with different distributions and even with very
low signal-to-noise ratios. (C) 1998 Pattern Recognition Society. Publ
ished by Elsevier Science Ltd. All rights reserved.