Claudius Jähn

Real-Time 3D Rendering of HeterogeneousźScenes

Many virtual 3D scenes, especially those that are large, are not structured evenly. For such heterogeneous data, there is no single algorithm that is able to render every scene type at each position fast and with the same high image quality. For a small set of scenes, this situation can be improved if different rendering algorithms are manually assigned to particular parts of the scene by an experienced user. We introduce the Multi-Algorithm-Rendering method. It automatically deploys different rendering algorithms simultaneously for a broad range of scene types. The method divides the scene into subregions and measures the behavior of different algorithms for each region in a preprocessing step. During runtime, this data is utilized to compute an estimate for the quality and running time of the available rendering algorithms from the observers point of view. By solving an optimizing problem, the image quality can be optimized by an assignment of algorithms to regions while keeping the frame rate almost constant.

Preprocessed global visibility for real-time rendering on low-end hardware

We present an approach for real-time rendering of complex 3D scenes consisting of millions of polygons on limited graphics hardware. In a preprocessing step, powerful hardware is used to gain fine granular global visibility information of a scene using an adaptive sampling algorithm. Additively the visual influence of each object on the eventual rendered image is estimated. This influence is used to select the most important objects to display in our approximative culling algorithm. After the visibility data is compressed to meet the storage capabilities of small devices, we achieve an interactive walkthrough of the Power Plant scene on a standard netbook with an integrated graphics chipset.

Evaluation of Rendering Algorithms Using Position-Dependent Scene Properties

In order to evaluate the efficiency of algorithms for real-time 3D rendering, different properties like rendering time, occluded triangles, or image quality, need to be investigated. Since these properties depend on the position of the camera, usually some camera path is chosen, along which the measurements are performed. As those measurements cover only a small part of the scene, this approach hardly allows drawing conclusions regarding the algorithms properties at arbitrary positions in the scene. The presented method allows the systematic and position-independent evaluation of rendering algorithms. It uses an adaptive sampling approach to approximate the distribution of a property like rendering time for all positions in the scene. This approximation can be visualized to produce an intuitive impression of the algorithms behavior or be statistically analyzed for objectively rating and comparing algorithms. We demonstrate our method by evaluating performance aspects of a known occlusion culling algorithm.

Algorithm Engineering Aspects of Real-Time Rendering Algorithms

Defining, measuring, and comparing the quality and efficiency of rendering algorithms in computer graphics is a demanding challenge: quality measures are often application specific and efficiency is strongly influenced by properties of the rendered scene and the used hardware. We survey the currently employed evaluation methods for the development process of rendering algorithms. Then, we present our PADrend framework, which supports systematic and flexible development, evaluation, adaptation, and comparison of rendering algorithms, and provides a comfortable and easy-to-use platform for developers of rendering algorithms. The system includes a new evaluation method to improve the objectivity of experimental evaluations of rendering algorithms.

Asynchronous Occlusion Culling on Heterogeneous PC Clusters for Distributed 3D Scenes

We present a parallel rendering system for heterogeneous PC clusters to visualize massive models. One single, powerful visualization node is supported by a group of backend nodes with weak graphics performance. While the visualization node renders the visible objects, the backend nodes asynchronously perform visibility tests and supply the front end with visible scene objects. The visualization node stores only currently visible objects in its memory, while the scene is distributed among the backend nodes’ memory without redundancy. To efficiently compute the occlusion tests in spite of that each backend node stores only a fraction of the original geometry, we complete the scene by adding highly simplified versions of the objects stored on other nodes. We test our system with 15 backend nodes. It is able to render a ≈ 350,M polygons (≈ 8.5,GiB) large aircraft model with 20, to 30,fps and thus allows a walk-through in real-time.