Modern industrial computed tomography can generate volume data of the
size of 1TB and larger and many of the scanned objects are unique, for
example cars in reverse engineering or cultural heritage.
To handle such ``bigtures'' on normal
hardware, compression is unavoidable and methods have to be carefully
designed that work on the sparse compressed data only. Once wavelet
compression is the tool of choice, this means that we can only use
the wavelet coefficients while access to the full decompressed image
is out of question. This affects all stages of the process, from
vizualization and standard image manipulations up to segmentation.

The talk presents some application examples and the mathematical
background especially for efficient real time denoising based on an
approximation of the TV norm from wavelet coefficients and a
semi-automatic segmentation process based on feature learning.