Hier finden Sie wissenschaftliche Publikationen aus den Fraunhofer-Instituten.

Fast automatic x-ray image processing by means of a new multistage filter for background modeling

: Hassler, U.; Heil, K.; Hanke, R.F.

Institute of Electrical and Electronics Engineers -IEEE-; IEEE Microwave Theory and Techniques Society:
ICIP 1994. International Conference on Image Processing. Proceedings. Vol.1
Los Alamitos: IEEE Computer Society Press, 1994
ISBN: 0-8186-6950-0
International Conference on Image Processing <1994, Austin>
Conference Paper
Fraunhofer IIS A ( IIS) ()
non-destructive testing; Qualitätssicherung; quality control; Röntgenbildverarbeitung; Röntgenprüfung; x-ray image processing system; x-ray inspection; zerstörungsfreie Prüfung

The automatic evaluation of radioscopic images includes the detection of deviations from the regular structure of the object under inspection. For this task two principle problems have to be taken into account: A way for powerful image segmentation must be found which is relative insensitive to noise because of the nature of x-ray images and the algorithm has to work very quick to satisfy industrial standard. Both problems touch each other because the design of the segmentation algorithm has to be a compromise between efficiency and reliability on the one side and processing time on the other side. The application of image processing methods in the field of non destructive testing like radioscopic inspection requires the consideration of some special conditions. While visual optical images usually contain sharp edges, less noise and relative high percentage of homogeneous greyvalue areas, x-ray images are highly contaminated by noise, have very smooth edges because of scattered radiati on and have only small homogeneous areas and many corner regions because of superposition; the latter condition srongly depends on complexity of the inspected object. This means that deviations or defects like flaws, pores or holes, which can be viewed as kind of impulsive noise, have to be detected both in edge and homogeneous noisy regions. In this paper the subsequent application of background modelling and pixel classification for the realization of this inspection task is presented.