Flexible Generation of Prior Images for Metal Artifact Reduction in Industrial Computed Tomography

Flexible Generation of Prior Images for Metal Artifact Reduction in Industrial Computed Tomography

Paper in proceedings outside WoS and Scopus

In computed tomography, artifacts caused by metal objects pose a major challenge, as they severely degrade the final image. Reduction of these artifacts is a topic of ongoing research, and many possible approaches have been developed so far. Among these, projection completion methods are effective and reasonably simple to implement compared to others. The most effective of these methods require a prior image, a synthetic, artifact-free representation of the data. This prior is used to recover data obscured by the metals and thus reduce the presence of potential secondary artifacts, but it may not be trivial to obtain especially in industrial computed tomography, where sample variability tends to be high. Traditional methods of prior generation are often based on multi-segmentation of a preliminary tomographic slice, which requires substantial a-priori knowledge about the scanned sample, its morphology, and the materials it is made of. This is a major limit for applications where such information is not readily available. Initial results for a more flexible method of prior image generation based on sinogram decomposition are presented here. The goal of this method is to decrease the need to fine-tune parameters for different samples and make it possible to use existing advanced metal artifact reduction methods in new fields.

In computed tomography, artifacts caused by metal objects pose a major challenge, as they severely degrade the final image. Reduction of these artifacts is a topic of ongoing research, and many possible approaches have been developed so far. Among these, projection completion methods are effective and reasonably simple to implement compared to others. The most effective of these methods require a prior image, a synthetic, artifact-free representation of the data. This prior is used to recover data obscured by the metals and thus reduce the presence of potential secondary artifacts, but it may not be trivial to obtain especially in industrial computed tomography, where sample variability tends to be high. Traditional methods of prior generation are often based on multi-segmentation of a preliminary tomographic slice, which requires substantial a-priori knowledge about the scanned sample, its morphology, and the materials it is made of. This is a major limit for applications where such information is not readily available. Initial results for a more flexible method of prior image generation based on sinogram decomposition are presented here. The goal of this method is to decrease the need to fine-tune parameters for different samples and make it possible to use existing advanced metal artifact reduction methods in new fields.

computed tomography, normalized metal artifact reduction, prior generation, sinogram decomposition

computed tomography, normalized metal artifact reduction, prior generation, sinogram decomposition

ZEMEK, M.; ŠALPLACHTA, J.; ZIKMUND, T.; KAISER, J.

2022

08.02.2022

NDT.net

Wels

11th Conference on Industrial Computed Tomography (iCT) 2022, 8-11 Feb, Wels, Austria

1

7

7

https://www.ndt.net/article/ctc2022/papers/ICT2022_paper_id255.pdf