Reconstructor 2.0
Fast scalable CUDA-based CT reconstructor for high resolution tomography (e.g. 10K × 3600 projections)
Customer:
Prodis.NDT (Russia)
Operating systems:
Windows, Linux

Technology stack:
C++ 14–17, CUDA, Python/C API
Software description

The upgraded version of Multi-GPU Reconstructor.

In this new version we implemented all the calculations using matrices, which made it possible to correct any geometric deviations (rotations, displacements, etc.). The structure of the software has become more modular. Performance and support for two GPUs have been also improved.

The new version allows automatic optimization of parameters and supports helical reconstruction. We have also developed a new algorithm for reconstructing parts of objects that go beyond the full coverage zone (that is, parts that go beyond the cylinder, as well as extreme upper and lower slices) to support reconstruction with incomplete data.

The reconstructor can be distributed in the format of a software integration kit with support for custom Python scripts. It was also implemented as a console application and included in the Viewer software.
Key features

  • Correction of any geometric deviations
  • Improved dual GPU support
  • Automatic parameter optimization
  • Helical reconstruction
  • Reconstruction with incomplete data
  • Support for custom Python scripts