3D tomographic reconstruction on HPC cluster of the Kongo Rikishi (Japanese wooden statue of the XIII century)

Abstract

Computed Tomography is a non-destructive technique for which the object volume is reconstructed from a large series of radiographs acquired at different angles. Information can be retrieved as 2D cross section images allowing the inspection and the classification of the object; moreover, by processing tomographic data, a 3D numerical model of the full-volume sample can be obtained for virtual reality applications or digital archives storage. Computed Tomography is well known, specially in the medical field. However, it is a complex technique as soon as one wants to use it in a different way than in medicine (that means different scale, different energy range, different material composition and so on). The Kongo Rikishi is a Japanese wooden statue (over 200 cm of height) of the XIII century. The restoration has been carried out by the Conservation and Restoration Center La Venaria Reale, Turin, Italy. An high resolution Computed Tomography has been realized by the X-Ray Imaging Group of the Physics Department of Bologna University, Italy. To investigate the whole statue volume, up to 36 shootings are needed and for each of them 720 radiographs are acquired. To obtain the final volume (120 GB) 20 days of calculation are needed with a standard PC. In this work we will show the Proof of Concept that we have done using the Microsoft HPC cluster in Redmond. The Microsoft HPC environment has proved to be dramatically powerful and easy to use letting us reach important results quickly. Simply running many copies of the same software on different chunks of data using 20 cores has lead to an impressive speedup: up to a speed factor of 28. The same code, running on a new generation cluster of 32 cores, completed the elaborations with a speed rate factor of 75. These extraordinary results permit to reconstruct the volume in 6 hours instead of 20 days, making possible the real-time investigation of cultural objects.