Georeferencing involves finding the mapping between the pixel coordinates of an image and world coordinates. Automating this process could save countless labor hours. I describe an automated algorithm (freely available at https://bitbucket.org/mpatmudd/opengeoreg) that combines (1) keypoint matching (e.g., SIFT or SURF) to determine the relative geometry between overlapping aerial images and (2) mutual information maximization (of pixel intensities) to determine the absolute geometry between aerial images and a reference image. I characterize the accuracy of the automated algorithm using a set of 32 historic aerial images from Australia's Western Desert. The algorithm performs as well or better than labor-intensive manual georeferencing.