Crack edge extraction and measure of industrial CT image based on ridgelet transform

Li Zeng, Beibei An, Taoyu Wan, Wentao Liu, P. R. Hunziker

Research output: Contribution to journalArticlepeer-review

2 Citations (Scopus)


The crack edge extraction is the basis of crack measure; however, it is not easy to be done by traditional methods because they are sensitive to noise in images. Accordingly, in this paper, a crack edge extraction method based on the ridgelet transform is presented to extract accurately crack edges of industrial computed tomography (CT) images from workpiece defect which can be detected by industrial CT. The ridgelet transform is realized by two steps, map the line singularity of an image into the point singularity of signals using the Radon transform, and detect the point singularity using wavelet transform. It performs well in detecting line singularities such as image edge, and it can restrain point noise of images. First, the direction and area of a crack are obtained by using the ridgelet transform. Then, the crack edge is extracted by area gradient operator and fitted to a polynomial, which is yielded a continuous crack edge. However, a deviation of the measured width of cracks from true width of real cracks may be caused by gray diffuse in CT images, so the width shrinkage and scale conversion was used in this case. Finally, the width of narrow cracks was measured accurately. As the accompanying experiments in this work indicate, this method provides not only a clearly delineated, continuous and unattached crack edge, but also sub-pixel accuracy for narrow crack width measurement. Copyright © 2009 Binary Information Press.

Original languageEnglish
Pages (from-to)1393-1401
Number of pages9
JournalJournal of Computational Information Systems
Issue number5
Publication statusPublished - Oct 2009


  • Crack measure
  • Edge extraction
  • Image processing
  • Industrial CT
  • Ridgelet transform
  • Sub-pixel


Dive into the research topics of 'Crack edge extraction and measure of industrial CT image based on ridgelet transform'. Together they form a unique fingerprint.

Cite this