Transient Motion Groups for Interactive Visualization of Time-Varying Point Clouds

Physical simulations provide a rich source of time-variant three-dimensional data. Unfortunately, the data generated from these types of simulation are often large in size and are thereby only experienced through pre-rendered movies from a fixed viewpoint. Rendering of large point cloud data sets is well understood, however the data requirements for rendering a sequence of such data sets grow linearly with the number of frames in the sequence. Both GPU memory and upload speed are limiting factors for interactive playback speeds. While previous techniques have shown the ability to reduce the storage sizes, the decompression speeds for these methods are shown to be too time and computation-intensive for interactive playback. This article presents a compression method which detects and describes group motion within the point clouds over the temporal domain. High compression rates are achieved through careful re-ordering of the points within the clouds and the implicit group movement information. The presented data structures enable efficient storage, fast decompression speeds and high rendering