An efficient method based on ABC for optimal multilevel thresholding

Abstract

Many efficient bi-level thresholding techniques have been proposed in recent years. Usually, the objective functions, which are used by them, are not appropriate for the multilevel thresholding owing to exponential growth of computational complexity. This work presents a new multilevel thresholding algorithm using Artificial Bee Colony algorithm (ABC) with the Otsu's objective function. Also, a strategy is used to guess suitable thresholds for initializing the proposed method. This initializing phase used the bi-level Otsu method to find the initial thresholds. These guessed thresholds are used to create a food source around each of them for use in the ABC algorithm as initial population. The presented thresholding method is tested on four popular images. The results show that this method has competitive performance compared to other well-known methods such as Gaussian-smoothing, Symmetry-duality, GA-based and PSO-based algorithms.