We describe an efficient approach to rendering a perspectively correct image on a potentially irregular display surface that may be illuminated with one or more distinct devices. The first pass of the technique generates an image of the desired graphics model using conventional rendering. The second pass projects that image as a texture onto a model of the display surface, then re-renders the textured display surface model from the viewpoint of each display device. The algorithm scales with the complexity of the display surface, and is constant with respect to the complexity of the graphics model.