Medical images aligning is essential due to the existing angular rotation and translation of objects inside the medical images which leads to complex post-processing. Accordingly, medical image registration becomes one of the key tools for aligning images. In image registration, to the images’ frames and the objects are transformed by aligning each image with its corresponding reference one. Demons registration is one of the most efficient non-rigid image registration techniques. In this paper, the proposed system optimized the parameters of Demons registration using the firefly optimization algorithm in retinal images. Several metrics were measured for evaluating the proposed model, including the mean square error, joint entropy, and mutual information calculation between the original and registered images.