Does anybody know of a link to a paper documenting how google maps (earth) works in terms of paging in higher resolution images when zooming in? I've tried to extrapolate the algorithm and I've only come up with a few ideas. The first is to start off with a single tile that contains the lowest resolution image. As the camera zooms in, the tile is evenly subdivided (2x2, 3x4, or however many times you want to subdivide it), and each subtile is loaded with a new higher solution image. As neighboring subtiles go out of the view frustum, they are paged out of memory. Then as the camera zooms into one of these subtiles, it is also subdivided into smaller tiles, each of which then gets a new high resolution image loaded. Each image that is loaded is always the same resolution, say 512x512. The only difference is obviously each image covers a smaller area as you zoom in. The second approach I came up with is a pre-constructed quadtree structure. First you could examine your dataset of images, determine where and at what level of zoom each should be loaded and then construct a quadtree based on this information. Hence, each node in the quadtree only gets pre-subdivided if there exists some higher resolution image within that particular area. Are any of these two assessments close to how google maps works? I'm attempting to create a project that mimics google maps but not with geographical images.