I finally downloaded OpenCV (an open source computer vision library) and spent some time learning about how to use it. It was actually surprisingly painless once I got the hang of it, and there are lots of things you can do once you feed it an image (in this case, the depth image). What I did was rig it to track contours on the depth image, and attempt to pick out a rectangular object.
Then, by using the detected location of the corners, I can apply it as a perspective transform to my laser projector. The end result is that the cardboard box I'm holding becomes a "virtual screen" that is tracked by the laser projection in real time and in perspective :)