Certain studies on thresholding based defect detection algorithms for Magnetic Flux Leakage images

Automatic thresholding is an important technique in the image segmentation process. The basic idea of automatic thresholding is to automatically select an optimal gray-level threshold value for partitioning pixels in the images into object and background based on their gray-level distribution. In this work the performance evaluation of three image thresholding algorithms namely, Otsu's thresholding method, Hou's thresholding method and Kapur's entropy based thresholding method for Non-Destructive Testing (NDT) applications were performed. Otsu's method is considered as one of the best threshold selection approaches for general real world images. The Otsu's and Hou's methods only consider class variance sum, but neglect variance discrepancy between foreground (object) and background. If the object is clearly distinguishable from the background, the gray-level histogram will be bimodal and the threshold for segmentation can be chosen at the bottom of the valley. Methods other than valley-seeking are thus required to solve this problem. So, Kapur's Entropy based thresholding method is performed for NDT images. This method uses global and objective property of the histogram. The suitable thresholding method for Magnetic Flux Leakage (MFL) image is identified by comparing performance measures for the three thresholding methods.