University of Konstanz
Graduiertenkolleg / PhD Program
Computer and Information Science

Graduation Talks

title

Geometry approximation in computer graphics and numerical wheather forecast

speaker

Tilo Ochotta, University Konstanz
Konstanz, Germany

date & place

Wednesday, 10.05.2006, 15:15 h
Room C252

abstract

The increasing demand for storing large amounts of 3D content requires compression methods that represent the data compactly. We present a generic framework for compression of densly sampled 3D-geometry with wavelet image compression methods. In our approach we decompose a given surface into a number of height fields that are resampled on regular grids over planar domains. The resulting images are encoded using a state-of-the-art compression method for shaped images. Our method is applicable to mesh- and point-based geometry and provides an excellent rate-distortion performance.

We further introduce a novel real-time point rendering pipeline that relies on components of state-of-the-art splat and mesh rendering methods. Utilizing the elevation map primitive, we implement an interpolative renderer that significantly improves the visual quality in comparison with the previous approaches.

In numerical weather prediction, measurements of various observation systems are combined with background data to define initial states for the forecasts. Current and future observation systems, in particular satellite instruments, produce large amounts of measurements with high spatial and temporal density. We propose two greedy thinning algorithms, which reduce the number of observations while retaining the essential information content of the data. We evaluated the proposed methods with respect to a geometric error measure and compared them to previous methods.