It works like google maps. The high resolution image is sampled into several layers of lower resolution. You can call it levels. Every level is split into quad

It works like google maps. The high resolution image is sampled into several layers of lower resolution. You can call it levels. Every level is split into quadrants. So when someone zooms in, it will fetch the quadrants that is visible to the browser viewport. By doing this way, the browser doesnt have to load the rest of the image and thus save alot of bandwidth!