Multi-Scale Structural Feature Extraction
Due to the technological improvements in computing and imaging devices exceedingly large volume data sets can be acquired e.g. with micro-CT (μCT) and Synchrotron Tomography (ST) scanning technology. These types of high-resolution 3D volume data sets significantly exceed what can be stored and managed easily in main (CPU) or graphics (GPU) memory. Accordingly, there is an ongoing need to analyze and adaptively reduce the amount of data prior to interactive rendering. A fundamental concept of data reduction is to remove redundant and irrelevant information while preserving the relevant features. Techniques of data reduction are thus directly linked to concepts and techniques of data compression, noise reduction and feature recognition/extraction. Therefore, we aim to capitalize on these links and introduce a novel tensor-approximation based framework to perform feature-adaptive data analysis and processing, and we will develop novel rendering methods based on this approach to visualize very large data sets interactively.