Compression and Reconstruction of Large Computed Tomography Data

In recent applications of computed tomography (CT), huge three-dimensional high-resolution data is generated. Thus, new methods are required to handle data exceeding the memory limitations of consumer hardware. In this talk, we focus on industrial CT data and describe a wavelet-based interactive (de)compression method allowing to efficiently store such huge datasets and additionally enabling a fast extraction of a local region of interest only, thus avoiding keeping the entire volume in main memory. Furthermore, we discuss a novel sparse reconstruction algorithm based on the Algebraic Reconstruction Technique which incorporates the ideas of the aforementioned compression method. This in turn has the benefit of directly producing a compressed volume with all its advantages while simultaneously circumventing the need for an uncompressed reconstruction followed by a compression.