The document discusses methods for obtaining a background image using depth information from a depth camera to more accurately extract foreground objects. It finds that accumulating depth images and taking the median value at each pixel provides the most accurate background image. The accuracy of three methods - average, median, and mode - are evaluated using simulated depth data of a captured plane. The median method provides the best results, followed by average, while mode performs worst. More accumulated images provide a more accurate background image across all methods.