The XYZ readings from the camera are in the reference frame of the camera. @FooBar is correct about the X/Y values: they are planar about the center of the camera, just like the OpenGL viewing window. I don't know the maximum range of the point-cloud data, but my suspicion is that the maximum z value is 1. (This could change, however, depending if you have a multi-camera setup, but I don't have the experimental dtaa to prove one way or the other, and that is outside the scope of the original question.)
RViz is built on OpenGL (though the robotics frames are not necessarily following the same standard, just that they share many similar design choices), so I suggest you look up the documentation for versions 4.X. Here is a great explanation of OpenGL coordinate frames, not that you needed to go into that much detail, but it helped me a lot.
Back to ROS: If you want to get the coordinates in the world frame rather than the local camera frame, there are 2 things you need.
- The location of the camera w.r.t. world frame
- The Point cloud data
You can then create a new frame transform object with no rotation offset from the camera (Unit quaternion with 0 rotations).
You can then take your camera frame and the newly computed frame and then extract the frame transform
- World frame
- Camera frame
- Point frame
Since I don't know what language you are using, I refer will refer you to these links regarding frame transforms.