Mostly, historical documents contain a lot of information which is more useful for different kinds of generation. They had been written for 1000 years ago, but these documents suffer from a lot of various types of noise that may be based on external factors such as poor storage, atmospheric factors as moisture, and high temperatures. In order to ensure the safety of these documents, they should be saved in a digital form to keep them safe, quick access to all their information easily using different available document analysis applications. In this paper, a global binarization approach for handwritten Arabic manuscript image binarization is proposed. This approach depends on employing a nature-inspired optimization algorithm called Moth-flames for minimizing K-means objective function. The used dataset consists of 50 handwritten manuscripts images. The proposed approach is compared with some of the well-known binarization approaches like Otsu's, and Niblack's. Experimental results were made in terms of different visual inspection, F-measure, p-FM, PSNR, GA, DRD, NRM, and MPM. Moreover, the comparison with the state-of-art methods proved the success of the proposed approach.