Weld defect features extraction on digital radiographic image using Chan-Vese model

Segmentation is the most important and critical task that widely used in image processing. In this study, Level set based on Chan-Vese model is explored and applied to detect weld defect on digital radiographic image. Region of interest (ROI) from original image is defined to reduce the area of image been processed. The ROI image is smoothen using median 3×3 filter to improve its quality for better detection result. Then, the defect in the ROI image is segmented using Chan-Vese model with two different size of rectangle as initial contour for initialization. The algorithm for the whole processes is implemented using MATLAB R2009a. From the detection result, its features are calculated that have been calibrated to absolute units (mm) from pixel units. The application of segmentation technique such as level set method able to assist radiographer in detecting and defining the properties of the defect accurately.