Tiled CT scans of large objects are typically performed with an overlap, that allows stitching of the individual scan tiles into a single volume. Regions of the complete CT scan are thus captured two times. In this project, a novel methodology for noise reduction in tiled CT scans will be developed. By training a convolutional neural network on noise-degraded images from the overlapped regions of the tiled CT scan, we can obtain an image noise reduction model, that can be used to improve the overall quality of the entire CT scan.

CEITEC VUT-J-22-8022

Czech

Matula Jan, Ing. - principal person responsible
Kaiser Jozef, prof. Ing., Ph.D. - fellow researcher

Advanced instrumentation and methods for material characterization
- responsible department (3.3.2022 - not assigned)
Central European Institute of Technology BUT
- responsible department (9.2.2022 - 3.3.2022)
Advanced instrumentation and methods for material characterization
- internal (1.1.2022 - 31.12.2022)
Central European Institute of Technology BUT
- beneficiary (1.1.2022 - 31.12.2022)

MATULA, J.; PELT, D.; VAN LEEUWEN, T.; ZIKMUND, T.; KAISER, J. Self-supervised learning for high quality tiled X-ray computed tomography imaging: a simulation study. Fifteenth International Conference on Machine Vision (ICMV 2022). Rome: 2022.
Detail

Responsibility: Matula Jan, Ing.