Elsevier

Neurocomputing

Volume 13, Issues 2–4, October 1996, Pages 261-279
Neurocomputing

Special paper
Practical applications of neural networks in texture analysis

https://doi.org/10.1016/0925-2312(95)00092-5Get rights and content

Abstract

In this paper we explore and discuss important issues concerning the use of neural networks in texture analysis. The advantage of integrating a neural network as a segregator in a modular approach for texture segmentation is explained. Then we show how to use a multilayer backpropagation neural network for texture classification and state its advantages and disadvantages. To avoid the disadvantages an approach based on the capability of a selforganizing-map neural network to approximate the probability density function of input data is introduced. This ability can be used to construct a Bayesian classifier based on trained selforganizing-maps and for a neural network equivalent to the analysis of variances (ANOVA) for defect detection on texturized surfaces. Examples of texture classification are given.

References (12)

  • F. Meyer et al.

    Morphological segmentation

    J. Visual Communication and Image Representation

    (1990)
  • R. Haralick et al.

    Image segmentation techniques

    Computer Vision, Graphics and Image Processing

    (1985)
  • P. Soille et al.

    Determining watersheds in digital pictures via flooding simulations

  • S. Beucher

    Watersheds of functions and picture segmentation

  • P. Soille et al.

    A modular system for image segmentation based on watershed transformation

  • J.F. Rivest et al.

    Morphological gradients

    J. Electronic Imaging

    (1993)
There are more references available in the full text version of this article.
View full text