Image geolocation, estimating GPS coordinates from an image, is a relatively new endeavor in the field of computer vision. This thesis presents two approaches to obtain the coordinates: hierarchical and dictionary-based. The hierarchical approach uses SVMs to first determine the general environment of the image and then estimates the exact location within that environment. The dictionary-based approaches are performed with linear and non-linear dictionaries using K-SVD and KK-SVD. Both methods are performed on the image feature gist and histograms of the image's color, SIFT descriptors, textons, and lines. Both the hierarchical and dictionary-based approaches build upon and combine existing systems to provide improved accuracy on a data set of twelve locations belonging to four environmental types.