IMOVIS: Interactive morphological and wavelet-based volume processing and visualization

IMOVIS poster at SIREN 2005

Summary:

Visualisation of large data sets requires advanced techniques in multidimensional signal processing, hierarchical data management, and data reduction. Data volumes generated by scientific simulations or produced by medical scanners can easily grow into the range of giga-bytes. Both the increasing size and complexity of these data drive the need for new techniques in interactive visualisation. The possibility of interaction during evaluation will significantly reduce the time required to interpret and present results. In this project, we propose to study the tight integration of volume processing and visualisation to obtain systems which can perform at interactive rates for very large data sets. In particular, we will systematically study multi-scale techniques from mathematical morphology and wavelet theory for reduction, feature extraction and real-time visualisation of volume data. By integrating these techniques in the visualisation pipeline in a highly automated fashion, the user is allowed to participate in the analysis. Also, we will study and extend particle systems for interactive feature extraction and segmentation based on physically motivated deformable models, and develop fast methods using tools from mathematical morphology, as an alternative to level-set and (nonlinear) diffusion approaches. For obtaining the required computational speed, we will implement these methods on modern programmable GPU hardware. The methods will be integrated as a software package using hardand software tools and libraries within existing open source visualisation and segmentation software environments. Evaluation of the effectiveness of our methods by both quantitative performance measures and user-studies will be an important part of this project.