At CoGVis/MMC, we believe that the interface paradigms used by current 3D modeling devices are such that their learning curve is too steep. This excludes from 3D modeling especially users which are good at arts and design but are not interested in learning technology. If 3D programs want to be diffused to all computer users, new interface paradigms have to be developed, which are self-explaining and understandable to all. We use the term natural interfaces to express this.
The real time simulation of physics is possible only since a few years. General purpose graphics card programming languages, as well as multicore CPUs allow to use part of their computing power to simulate physics in real time. The real time simulation of physics is a particularly challenging problem: while offline physics simulations can compensate instability issues through finer sampling on the time axis, this is in general not possible if real time display is required. Commonly used finite element methods suffer under real time requirements of numerical instability. Moreover, the tradeoff between computing speed of the numerical solver and stability is particularly difficult to find.
Photorealistic Computer Graphics encompasses the rendering of complex materials which are present in nature or are manufactured by craftsmen or industry. Simulating lighting has been at the center of graphics research in the past three decades. Rendering equations simulate light reflecting properties of the surface of an object. Rendering equations are useless unless reasonable and detailed material data is acquired and made ready for the rendering process. There is a quite big deficit on material data, because data acquisition has been made mainly on behalf of industry, and such data is not freely available. At CoGVis/MMC we are working on hardware and methods for a low cost automatic acquisition of material properties for rendering purposes. This includes the automatic acquisitions of Bidirectional Texture Functions (BTFs) and the rendering of the acquired data in real time.
Selected Projects
TreeHugger (Bernhard Bittorf)
The project aims at generating, simulating and visualizing trees.
topics: growth (L-Systems vs. Particels); Modelling (Generalized Cylinders, NURBS?); Visualisierung (OpenGL, GLSL, OpenCL, Displacement Maps, Shadows); Physics (Rigid-body vs. Euler-Bernoulli-Beam-Model); Simulation of 3D windfields (Navier-Stokes); Light calculation (Raycasting vs. atomic kitten); Pruning of dynamically generated trees; Procedural growth dependent growth rings; Procedural aging wine-leafes; Synthesizing bark using real-time tesselation
Dromedar (Bernhard Bittorf)
This is a collaboration with the German Archaeological Institute Cairo and the Faculty of Architecture. The great archeological excavation sites are to be brought to life.
Perception and Analysis of Quality in Images (Prof. Dr. Charles A. Wüthrich and M. Sc. Banafsheh Azari)
There are many - relatively simple - methods for evaluating of image quality. In the field of image acquisition new methods have recently surfaced, allowing a more detailed view of various aspects of quality, both at the micro- and macro-level. The project will focus on understanding what defines quality and what factors are perceived as quality in imaging and rendering. When humans look at static pictures, where do they look? When humans look at computer generated pictures, where do they look at? Do they explore the same picture differently if they look at the same image rendered at different quality levels? Is resolution important? When is detail important? These and similar open questions will be central themes of the project.