Progressive Compression of Normal Vectors

We present a method for compression of normal vectors that (1) is general, in the sense that it can be incorporated into many different progressive geometry compression schemes; (2) is progressive, and accommodates arbitrary view dependent refinement; and (3) gives compression ratios and rate-distortion performance superior to previous methods. Our method works by progressively refining a normal cone to one of eight possible sub-cones (including no refinement), and entropy encoding the choice. We experimentally demonstrate the effectiveness of this approach on both meshes and point clouds, in combination with two very different geometry progression methods.