||Visualizing large, multiply-attributed, and time-dependent graphs is one of the grand challenges of information visualization. In recent years, image-based techniques have emerged as a strong competi- tor in the arena of solutions for this task. While many papers on this topic have been published, the precise advantages and limitations of such techniques, and also how they relate to similar techniques in the more traditional fields of scientific visualization (scivis) and image processing, have not been sufficiently outlined. In this paper, we aim to provide such an overview and comparison. We highlight the main advantages of image-based graph visualization and propose a simple taxonomy for such techniques. Next, we highlight the differences between graph and scivis/image datasets that lead to limitations of current image-based graph visualization techniques. Finally, we consider these limitations to propose a number of future work directions for extending the effective- ness and range of image-based graph visualization.