Understanding the Kinect pointcloud

edit

Hi,

I saved the pointclouds published by kinect in a pcd file. Width is 640 and height is 480.

I expect kinect to give the x, y, z for every pixel in 640x480.

I believe that the sequence of the pointcloud data starts from the left bottom of the image to the right top of the image, pixel wise left to right.

I understand some of the values are nan as points at that point are not defined as they are not visible or due to some noise.

The z component almost gives the distance of the object from the kinect camera.

I don't understand the values of x y in the point cloud. or they normalized with respect to something, if so how to make it global? Is my understanding wrong and they are still the exact distance in meters? I thought kinect drivers are built for ros so its right to pose this question here.

Thanks, Karthik