Detection threshold value by using optimization algorithms
Sara Fadhil, Kadhim Mahdi Hashim · AIP conference proceedings · 2022
Choosing optimal threshold value is a very important step in image segmentation. In this paper, we suggest a new method for image thresholding by using an optimization algorithm to detect a good threshold value that separates the image into foreground and background. In the image which is bi-model histogram contain two peaks , it needs one threshold to separate the object about the background when the histogram became complex more than two peaks needs more than one value a Multilevel Threshold (MT). For the classical techniques of obtaining threshold values, their efficiency and accuracy have been already proved for a bi-level segmentation. Although they can be expanded for MT, their computational complexity increases exponentially when a new threshold is obtained. The MT problem has also been handled through evolutionary optimization methods. In general, they have demonstrated to deliver better results than those based on classical techniques in terms accuracy, speed and robustness. Proposed method used two optimization algorithm harmony search algorithm (HS)and electromagnetic field optimization(EFo) to detect optimal threshold values and used for segmentation. These algorithms select random samples from the histogram of image as search space and used Kapur’s method as objective function to measure the goodness of these values. Some of misclassification measurement s used to compare the result segmented image with it’s ground truth from Berkeley data set .Another performance measurement used to evaluate the proposed methods for image segmentation