NDTnet 1998 Aug, Vol.3 No.8

Abstract

Introduction
Non Destructive Testing of materials has to allow to obtain the highest possible probability detection, the most just size and the exact orientation of dangerous defects that the specimen to control can contain. In recent years, the power of computers has enabled the emergence of sophisticated algorithms dedicated to process and interpret rapidly a huge amount of information (especially digital image processing ).In this paper, segmentation method based on the analysis by Co-Occurrence Matrix is developed. It is an effective tool that has been introduced by HARALICK. Co-Occurrence Matrix has been withheld among the various methods of segmentation for their selective analysis abilities of the image following a direction of observation and by the possibility to regard as the noise of the analysed image as a texture. These two points are perfectly adapted to ultrasonic images (1). We apply also this tool to X-rays images that are currently increasingly studied. We are interested especially in welded joint radiograms so as to make a contribution to a detection and an automatic failure interpretation..

Fig. 1. Example of Co-Occurrence Matrix coded in 256 grey levels. Fig.2. B-Scan image. Fig.3. B-Scan Segmented.

Co-Occurrence Matrix and Results
In this work, we opt for the general definition of this matrix using spatial gray level dependence method. We study spatial and angular relations between back-ground and defects.The Co-Occurrence Matrix is a function of two variables i and j, the intensities of two pixels, it takes its elements in N. The parameters of this matrix are the image f and the vectord. Elements of matrix represent the number of occurrenceon the image f, of pairs of pixels separated by vector d.

C: E x E -> N
(i,j) -> c( i, j, f, d)
We use two methods:


• Manual thresholding by detecting peaks of matrix diagonal to segment X-Rays images.
  
• Automatic thresholding by dividing the matrix in several blocks to segment B-Scan images.Practical results (fig. 1,2,3) are presented and commented

  References :
  1. G. Corneloup, J. Moysan & E. Magnin., B-Scan image segmentation by thresholging usingcooccurrence Matrix analysis., Pattern Recognition. 1996, volume 29. N°2, pp281-296

Abstract Source:
Book of Abstracts, 7th European Conference on Non-Destructive Testing, 26-29 May 1998, ISBN: 87-986898-0-00

Full-Text Source:
Proceedings of the 7th European Conference on Non-Destructive Testing, 26-29 May 1998, ISBN:

Publication Contact:
7th ECNDT, Park Alle 345, DK-2605 Broendby, Denmark, Fax: +45 46 26 70 11, Email: 101373.3414@compuserve.com