DefinePK

Abstract

Image thresholding is most important in image processing for unfolding foreground objects. In recent years, many thresholding methods have been proposed. However, identifying the weld defects on weld X-radiography images is a challenging task in Non-Destructive Testing (NDT) methods. Generally radiography images are in low contrast and lack of details. It is very difficult to extract the weld defects that are present. The main goal on this paper is to give comparative discussion on different threshold-based segmentation methods through performance measures. The thresholding techniques applied on various weld X-radiography images and their performance have been evaluated by non-uniformity (NU), misclassification error (ME) and relative foreground area error (RAE) measures. This performance analysis is supportive for an appropriate use of existing thresholding techniques on weld X-radiography image segmentation.