Performance Measurement of Thresholding Algorithms in Printed Circuit Board Inspection System

It is well-known that thresholding algorithm is needed by gray-scale images in segmentation operation. In this paper, real printed circuit boards used are captured using a monochrome charged coupled device camera and produced images consisting of foreground and background objects in gray-scale format. In order to correctly differentiate these two kinds of attributes, several of thresholding algorithms are applied. Then, misclassification error measure has been employed to evaluate the performances of these thresholding algorithms. This method gives information on how many percentages of background pixels wrongly assigned to foreground, and foreground pixels wrongly assigned to background pixels. As the best thresholding algorithm has been chosen, median filtering and the proposed defect classification algorithm can be applied to remove small noise in resultant images and classify the defects occurred on the printed circuit boards, respectively.