Smooth visualization of large point clouds

We present a novel approach for processing and rendering large point cloud data from 3D scanners on a standard computer system. Our technique handles a data size that typically exceeds the RAM for processing and the VRAM for rendering the data. We achieve interactive framerates and a smooth visualization by enhancing existing large data visualization techniques, like Level of Detail and deferred rendering approaches and we utilize the new capabilities of modern GPU hardware. This enables the exploration of large and unstructured 3d scan data sets whose size is only limited by the hard disk memory available. We verify our approach at the example of different data sets with up to one billion 3D points.