Segmentation of X-ray and C-scan images of fiber reinforced composite materials

Abstract

This paper is concerned with the segmentation of images corrupted by non-uniform illumination and sensor noise. Four segmentation techniques, i.e. simple thresholding, the adaptive thresholding scheme of Chow and Kaneko, the iterated conditional modes method of Besag, and the adaptive thresholding scheme of Yanowitz and Bruckstein, are presented here. The results of these four methods applied to three sets of artificial and real images are compared. Since these segmentation results are not satisfactory, an algorithm based on the adaptive thresholding method proposed by Yanowitz and Bruckstein and the Canny edge detector is proposed for the segmentation of real X-ray and C-scan images of composite materials. This method has been applied to several images of composite specimens containing a variety of defects. The segmentation results of X-ray and C-scan images are satisfactory which are further improved by a simple fusion technique.