Pelt, D; Maughan-Jones, C; Roche I Morgo, O; Olivo, A; Hagen, D; (2020) Rapid and flexible high-resolution scanning enabled by cycloidal computed tomography and convolutional neural network (CNN) based data recovery. In: Proceedings of the 6th International Conference on Image Formation in X-Ray Computed Tomography.

Preview

Text Abstract_PeltHagen_final_revised.pdf - Accepted Version Download (487kB) | Preview

Abstract

We have combined a recently developed imaging concept (“cycloidal computed tomography”) with convolutional neural network (CNN) based data recovery. The imaging concept is enabled by exploiting, in synergy, the benefits of probing the sample with a structured x-ray beam and applying a cycloidal acquisition scheme by which the sample is simultaneously rotated and laterally translated. The beam structuring provides a means of increasing the in-slice spatial resolution in tomographic images irrespective of the blur imposed by the x-ray source and detector, while the “roto-translation” sampling allows for rapid scanning. Data recovery based on the recently proposed Mixed-Scale Dense (MSD) CNN architecture enables an efficient reconstruction of high-quality, high-resolution images despite the fact that cycloidal computed tomography data are highly incomplete. In the following, we review the basic principles underpinning cycloidal computed tomography, introduce the CNN based data recovery method and discuss the benefit of combining both.

Download activity - last month

Export as JSONXMLCSV

Download activity - last 12 months

Export as CSVXMLJSON

Downloads by country - last 12 months

Export as CSVXMLJSON

Archive Staff Only

View Item