@misc{LamasRodriguezEhlkeHoffmannetal., author = {Lamas-Rodr{\´i}guez, Juli{\´a}n and Ehlke, Moritz and Hoffmann, Ren{\´e} and Zachow, Stefan}, title = {GPU-accelerated denoising of large tomographic data sets with low SNR}, issn = {1438-0064}, url = {http://nbn-resolving.de/urn:nbn:de:0297-zib-56339}, abstract = {Enhancements in tomographic imaging techniques facilitate non-destructive methods for visualizing fossil structures. However, to penetrate dense materials such as sediments or pyrites, image acquisition is typically performed with high beam energy and very sensitive image intensifiers, leading to artifacts and noise in the acquired data. The analysis of delicate fossil structures requires the images to be captured in maximum resolution, resulting in large data sets of several giga bytes (GB) in size. Since the structural information of interest is often almost in the same spatial range as artifacts and noise, image processing and segmentation algorithms have to cope with a very low signal-to-noise ratio (SNR). Within this report we present a study on the performance of a collection of denoising algorithms applied to a very noisy fossil dataset. The study shows that a non-local means (NLM) filter, in case it is properly configured, is able to remove a considerable amount of noise while preserving most of the structural information of interest. Based on the results of this study, we developed a software tool within ZIBAmira that denoises large tomographic datasets using an adaptive, GPU-accelerated NLM filter. With the help of our implementation a user can interactively configure the filter's parameters and thus its effectiveness with respect to the data of interest, while the filtering response is instantly visualized for a preselected region of interest (ROI). Our implementation efficiently denoises even large fossil datasets in a reasonable amount of time.}, language = {en} } @article{LamasRodriguezHerasArgueelloetal.2013, author = {Lamas-Rodr{\´i}guez, Juli{\´a}n and Heras, Dora Blanco and Arg{\"u}ello, Francisco and Kainm{\"u}ller, Dagmar and Zachow, Stefan and B{\´o}o, Montserrat}, title = {GPU-accelerated level-set segmentation}, series = {Journal of Real-Time Image Processing}, journal = {Journal of Real-Time Image Processing}, publisher = {Springer Berlin Heidelberg}, issn = {1861-8200}, doi = {10.1007/s11554-013-0378-6}, pages = {1 -- 15}, year = {2013}, language = {en} }