Techniques for real-time high quality rendering of complex models

Despite recent advances in finding effective LOD-Representations for gigantic 3D objects, rendering of complex, gigabyte-sized models and environments is still a challenging task, especially under real-time constraints and high demands on the visual accuracy. In the first part of this talk I will give an overview over our recent results on the simplification and efficient hybrid rendering of complex meshes and point clouds. After introducing the general hierarchical concept I will present two hybrid LOD algorithms for real-time rendering of complex models and environments. In the first approach we use points and triangles as the basic rendering primitives. To preserve the appearance of an object a special error measure for simplification was developed which allows us to steer the LOD generation in such a way that the geometric as well as the appearance deviation is bounded in image space. A novel hierarchical approach supports the efficient computation of the Hausdorff distance between the simplified and original mesh during simplification. In the second approach we refrain from using triangles in combination with points. Instead we replace most of the points by planes. Using these planes the filtering and therefore the rendering quality is comparable to elaborate point rendering methods but significantly faster since it is supported in hardware. In the second part we concentrate on efficient GPU based rendering of Trimmed Non-Uniform Rational B-Spline surfaces (NURBS). Due to the irregular mesh data structures required for trimming there were no algorithms that exploit the GPU for tessellation so far. Instead, all recent approaches perform a pre-tessellation and use level-of-detail techniques in order to deal with complex Trimmed NURBS models. In contrast to a simple API these methods require tedious preparation of the models before rendering. In addition this pre-processing hinders interactive editing. With our new method the trimming region can be defined by a trim-texture that is dynamically adapted to the required resolution and allows for an efficient trimming of surfaces on the GPU. Combing this new method with a GPU-based tessellation of cubic rational surfaces allows a new rendering algorithm for arbitrary trimmed NURBS and even T-Spline surfaces with prescribed error in screen space on the GPU. The performance exceeds current CPU-based techniques by a factor of about 200 and makes real-time visualization of trimmed NURBS and T-Spline surfaces possible on consumer-level graphics cards.