Abstract: Image thresholding was the process of converting grayscale or even color images into images that had fewer classes of possible pixel values. Thresholding methods could involve finding either a single threshold value (bi-level) or multiple thresholds (multilevel). Bi-level thresholding method was straightforward, but multilevel methods involved exhaustive searching that required large amounts of computation time. One meta-heuristic optimization method to solve the computation time problem was based on bee’s behavior in nature. The recently introduced variant of this method was the Bees Algorithm (BA). BA mimics honey bee foraging activities. It had been proven to be the most powerful fair optimization method for sampling a large solution space because of its fair random sampling. In this study, Otsu’s BA-based method was used to reduce computation time in multilevel image thresholding. Two standard images, Lena and Peppers, were thresholded using the peak signal-to-noise ratio as the image quality index. The effectiveness of the proposed method in terms of its peak signal noise ratio and computation time was measured. The results were then benchmarked against other optimization algorithms, such as Artificial Bee Colony (ABC), Honey Bee Mating Optimization (HBMO), Particle Swarm Optimization (PSO) and excessive search. The experiments showed that the quality of images generated by the BA was the best among all of the methods. The BA also used the shortest computation time to find more than 4 thresholds. This result demonstrates that the BA was an outstanding method for optimizing multilevel image thresholding, especially for large threshold values.