Anton Marini, a.k.a Vade, has made a beta (or alpha?) plugin that delivers the depth map output from the Kinect, which has substantially improved the quality of mesh or point clouds generated from the depth map, since it delivers an intensity image.
You can "see" how the ridges of depth are more tightly packed in these two example pics. It requires some changes to my code, since luminosity doesn't=near Z, and the image output type doesn't get handled automatically by the openCL kernel. At the end of the day, it does work with GLSL and OpenCL though.