Restoration and Segmentation of Rail Surface Images

  • Thomas H. Short
Conference paper
Part of the Lecture Notes in Statistics book series (LNS, volume 83)


Automated detection and classification of rail surface defects would be of great benefit to railroad companies. However, video images of the rail surface are subject to various forms of distortion. In order to accurately locate and measure defects, the images must be restored to an estimate of their ideal, undistorted form.

Two algorithms are suggested for the simultaneous restoration and segmentation of objects in a two-dimensional image. Both algorithms rely on distributions that model the relationship between the intensity of a pixel and the intensities of the pixel’s neighbors, to produce a restored version of the image.

The Iterated Conditional Averages algorithm (Johnson 1989) is a method for image restoration and edge detection. Examination of the estimated edge site values leads to a possible classification scheme for detecting defective sections of rail.

The Iterated Conditional Modes procedure for image restoration (Besag 1986) is extended to also perform object location and measurement. It is capable of providing detailed measurements of the geometric features of objects detected in an image.

Examples of the application of both algorithms to videotaped images of the surface of railroad track are presented. The detection of defects on the rail surface was the primary motivation for this research, and the algorithms are evaluated based on their potential ability to recognize and classify images containing sections of defective rail.


Image Processing Iterated Conditional Averages Iterated Conditional Modes 


Unable to display preview. Download preview PDF.

Unable to display preview. Download preview PDF.


  1. Alfelor, R. (1991) Analysis of automatically collected rail surface defect data for rail maintenance management, Ph.D. Dissertation, Department of Civil Engineering, Carnegie Mellon University.Google Scholar
  2. Besag, J. (1986) On the statistical analysis of dirty pictures, Journal of the Royal Statistical Society, Ser. B, 48, pp. 259–302.MathSciNetzbMATHGoogle Scholar
  3. Geman, S. and Geman, D. (1984) Stochastic relaxation, Gibbs distributions, and the Bayesian restoration of images, IEEE Transactions on Pattern Analysis and Machine Intelligence, 6, pp. 721–741.zbMATHCrossRefGoogle Scholar
  4. Johnson, V. (1989) On statistical image reconstruction, Ph.D. Dissertation, Department of Statistics, The University of Chicago.Google Scholar
  5. McNeil, S., Motazed, B., Alfelor, R., and Short, T. (1991) Automated railhead surface inspection, International Advances in Nondestructive Testing, McGonnagle, W.J. (ed.), Vol. 16, Philadelphia: Gordon and Breach.Google Scholar
  6. Short, T.H. (1991) Algorithms for simultaneous image restoration and segmentation, Ph.D. Dissertation, Department of Statistics, Carnegie Mellon University.Google Scholar

Copyright information

© Springer-Verlag New York, Inc. 1993

Authors and Affiliations

  • Thomas H. Short
    • 1
  1. 1.Villanova UniversityUSA

Personalised recommendations