TOP-CT: Trajectory with Overlapping Projections X-ray Computed Tomography
IEEE Transactions on Computational Imaging , Volume 8 p. 598- 608
TOP-CT (Trajectory with Overlapping Projections X-ray Computed Tomography) is a new class of CT scanning geometries for high throughput industrial CT scanning. In TOP-CT multiple objects move with a constant spacing over the same trajectory between a stationary X-ray source and detector. The projections of multiple objects can overlap, which provides additional flexibility when designing CT scanning geometries. Reconstruction algorithms were developed to reconstruct objects one by one from the overlapping projection data as soon as the objects move out of the field of view of the scanning setup. This makes it possible to make reconstructions while new objects with overlapping projections keep being added. The forward problem of TOP-CT is linear with a band block Toeplitz structure, and the matrix of the forward problem can be constructed from multiple copies of a non-overlapping CT projection matrix, so existing software toolkits can be used for TOP-CT with only a small modification. Simulation experiments and a real life experiment were performed on a U-turn TOP-CT geometry. One experiment showed that reconstructions from an overlapping projection setup have a slightly higher SSIM (0.828 vs 0.811) and similar PSNR (33.50 vs 33.34) compared to a non-overlapping setup, using the same scan time per object and the same reconstruction algorithm (SIRT). Another experiment showed that a reconstruction algorithm making reconstructions one by one using only local projection data performed without loss of quality compared to a baseline reconstruction method using all projection data.
|, , , , , ,|
|IEEE Transactions on Computational Imaging|
|Universal Three-dimensiOnal Passport for process Individualization in Agriculture|
Schut, D.E, Batenburg, K.J, van Liere, R, & van Leeuwen, T. (2022). TOP-CT: Trajectory with Overlapping Projections X-ray Computed Tomography. IEEE Transactions on Computational Imaging, 8, 598–608. doi:10.1109/TCI.2022.3192125