Image-based rendering (IBR) is unique in that it requires computer graphics, computer vision, and image processing to join forces to solve a common goal, namely photorealistic rendering through the use of images. IBR as an area of research has been around for about ten years, and substantial progress has been achieved in effectively capturing, representing, and rendering scenes. In this article, we survey the techniques used in IBR. Our survey shows that representations and rendering techniques can differ radically, depending on design decisions related to ease of capture, use of geometry, accuracy of geometry (if used), number and distribution of source images, degrees of freedom for virtual navigation, and expected scene complexity.