Proseminar "A Trip Through Computer Graphics" SS 2009
Downloads
Downloads only work from the RWTH, BIT or VPN network. For information on VPN see the RZ website.
Events
26th Feb |
|
|
| #6317 |
|
|
| Kickoff-Meeting |
2nd Mar, 3rd Mar, | CS Library, | Library Course | ||||||
3rd Jul | Preliminary paper due | |||||||
27th Jul | Final paper due | |||||||
10th-14th Aug | Optional test-presentation
| |||||||
17 Aug & 19 Aug
| #6317 | Final presentation |
Requirements
You will have to prepare both a short report (about 8 pages) and a talk of about 25 minutes. The report is supposed to be written in English, but can optionally be German. The talk can be in English, too, which is recommended as a practice for talks which you will hold later in your studies.
The goal of a proseminar is to learn to acquire new information from scientific publications on your own, to summarize them in your own words and to learn how to give a talk about this topic. The main focus is on learning the form and methodology, and the content is secondary. The form and methodology include:
- Report
- Literature research (including a short course at the library)
- Writing with the LaTeX wordprocessing system
- Correct literature referencing
- Presentation
- Creation of slides
- Abstraction and summarizing a topic
- Delivering a topic to an audience
Content
In computer graphics we use algorithms to generate images synthetically and to integrate or alter visual and spatial information sampled from the real world. This proseminar provides an insight into all important areas of computer graphics. In the following we show some aspects of computer graphics our current projects are involved in:
Using our 3D-scanner we are able to generate point clouds of arbitrary real objects (sampling of the real world). These clouds need to be tesselated for further processing, i.e. we have to create a mesh (usually a triangle mesh) out of the points.
Real data, as provided by the 3D-scanner, is noisy in contrast to synthetic data. There are different approaches for smoothing and improving the visual appearance.
Another way for improving visual appearance are subdivision algorithms. These allow successive refinement of the coarse meshes leading to coarse-to-fine hierarchies. Using mathematical tools, subdivision algorithms are analysed. The convergence analysis provides information about the smoothness of the limit surface.
Not only coarse-to-fine but also fine-to-coarse hierarchies of the meshes can be very useful, e.g. for progressive data transmission. If meshes become too complex for realtime-processing they have to be decimated using some appropriate algorithm. During the decimation process we especially have to take care of the resulting error that is made in comparison with the original mesh.
Handling large data sets implies usage of very efficient data structures, which enable us to control level-of-detail of the rendered meshes. During rendering passes only parts of the scene can be chosen to be shown at some fine level, leaving the rest rather coarse and thus leading to a higher rendering performance.
Rendering is an interesting and complex topic for itself. After acquiring and processing the mesh data, we are able to render the geometry with different algorithms. Depending on quality and speed we can visualize data in realtime, or we can use elaborated raytracing algorithms to get the highest possible quality.
Possible topics
- Affine and projective geometry
- Raster graphics
- Color spaces
- Graphics APIs (OpenGL, etc.)
- Scenegraph APIs
- Illumination and rendering
- Shading
- Texture mapping
- Shadows
- Global Illumination
- Volume Rendering
- Geometry representation
- Curves and surfaces
- Triangle meshes
- Point based geometry
- etc.