Computer graphics & visualization

Over the past few decades computer graphics has evolved into a broad field covering diverse areas such as image synthesis (rendering), image processing, geometry processing, texturing, animation, object reconstruction, perception, physics simulation, material appearance modelling, and computer aided design.

The applications are plenty and diverse, ranging from computer games over medical visualization to architectural design. In the Image Analysis and Computer Graphics Section at DTU Compute, our graphics research is focused on rendering and geometry processing.

   

Reliable Surface Reconstruction from Deformable 3D Scans

The focus is on reconstructing sequences of coherent deformable range images, acquired with high-frame rate capturing machines from bodies that move and deform, into plausibly watertight surfaces for the desired time slots. The focus is on the interaction between geometry and topology along with an efficient treatment of singularities

 

Global Illumination

 

The richness of real world illumination is rarely captured in computer graphics. However, many inroads have been made in recent years, and it is now possible to render effects such as reflections, shadows, indirect illumination, caustics, translucent materials and participating media.

Most of these effects have been made possible through the use of Photon Mapping. Photon mapping was originally invented at the department and it is still an area of research. However, many global illumination effects are still hard or impossible to achieve in real-time rendering. For instance, soft shadows in real-time systems is still a research topic. Hence, another important focus of our research is to incorporate global illumination effects in real-time rendering.

Ultimately, the goal is to blur the distinction between real-time rendering and off-line rendering. Recent real-time projects include multiple specular reflections from curved and planar reflectors, accelerated form factor computation for radiosity, dynamic relighting, and soft shadows. Read more about this here...

For further information, please contact: Niels Jørgen Christensen, njc@imm.dtu.dk

 

Intuitive, Interactive Shape Modelling

Interactive modelling of 3D shapes on a computer should be as simple and intuitive as doodling 2D shapes using pencil and paper. Simpler, in fact, since on a computer changes can always be undone, and the user is more free to explore and experiment. Unfortunately, completely intuitive, interactive modelling of shapes seems to be an elusive goal.

Some researchers attack the problem of intuitive sculpting from a user interface perspective. Perhaps the best example is the well-known gesture-based system, Teddy, by Takeo Igarashi. It seems likely that by a careful design of user friendly interfaces, we can come a long way towards intuitive sculpting, but, unfortunately, it also seems that the underlying representation will be an obstacle - simply because effective modelling often requires the user to be aware of the parameters of the representation.

The goal of this project is to explore new methods for shape modelling that allow the user to be completely unaware of the underlying representation so that she can focus solely on the task of modelling. Inroads have been made, using the volume representation in conjunction with the Level-Set Method. Future challenges involve handling high resolution detail and adaptive resolution. We are also looking into the application of shape modelling techniques to surgery planning. This is also a part of the "Visualization and Manipulation of Medical Data" project discussed elsewhere. Read more about this here...

Contact: Andreas Baerentzen, jab@imm.dtu.dk 

 

 

Geometric Modelling

 

Geometric modelling is a broad topic. Many algorithms have been designed to support CAD software, and one line of research in this department is the development of new, intuitive techniques for geometric or shape modelling (or sculpting).

In recent years, many geometry processing algorithms have been developed in order to support geometry produced by various types of scanning such as laser scanning, structured light scanners, time of fligh cameras, structure from motion and also surfaces obtained by tesselating isosurfaces in CT or MR scans of (typically) organic or medical objects. Our research interests include the ability to remove noise from such models, deform shapes naturally, and to develop robust algorithms which can handle data with e.g. topological noise. In many cases, it is necessary to convert between various shape representations such as triangles meshes, distance fields, point sets, functional representations such as radial basis functions or tetrahedral complexes.

For further information, please contact: Andreas Baerentzen, jab@imm.dtu.dk 

 

 

Visualization and Manipulation of Medical Data

 

Medical visualization on commodity hardware has become feasible thanks to the enormous recent advances in the field of consumer graphics cards. This project is a part of the collaborative 3D-Med project whose goal is the development of a PC-based medical workstation for viewing and measurement of volumetric, medical data such as CT or MR scans.

Our goals include fast texture based visualization of large volumes, flexible and intuitive tools for adjusting transfer functions, likewise intuitive tools for measurement, and finally tools for manipulation of bone structures using shape modelling techniques.

For further information, please contact: Niels Jørgen Christensennjc@imm.dtu.dkAndreas Baerentzen, jab@imm.dtu.dk

Contact

Jakob Andreas Bærentzen
Associate professor
DTU Compute
+45 45 25 34 14

Contact

Niels Jørgen Christensen
Associate Professor
DTU Compute
+45 45 25 33 66