TY - GEN
T1 - Beyond Local Processing
T2 - 17th European Conference on Computer Vision, ECCV 2022
AU - Hamoud, Bassel
AU - Bahat, Yuval
AU - Michaeli, Tomer
N1 - Publisher Copyright:
© 2023, The Author(s), under exclusive license to Springer Nature Switzerland AG.
PY - 2023
Y1 - 2023
N2 - Convolutional neural networks (CNNs) are well suited for image restoration tasks, like super resolution, deblurring, and denoising, in which the information required for restoring each pixel is mostly concentrated in a small neighborhood around it in the degraded image. However, they are less natural for highly non-local reconstruction problems, such as computed tomography (CT). To date, this incompatibility has been partially circumvented by using CNNs with very large receptive fields. Here, we propose an alternative approach, which relies on the rearrangement of the CT projection measurements along the CNN’s 3rd (channels’) dimension. This leads to a more local inverse problem, which is suitable for CNNs. We demonstrate our approach on sparse-view and limited-view CT, and show that it significantly improves reconstruction accuracy for any given network model. This allows achieving the same level of accuracy with significantly smaller models, and thus induces shorter training and inference times.
AB - Convolutional neural networks (CNNs) are well suited for image restoration tasks, like super resolution, deblurring, and denoising, in which the information required for restoring each pixel is mostly concentrated in a small neighborhood around it in the degraded image. However, they are less natural for highly non-local reconstruction problems, such as computed tomography (CT). To date, this incompatibility has been partially circumvented by using CNNs with very large receptive fields. Here, we propose an alternative approach, which relies on the rearrangement of the CT projection measurements along the CNN’s 3rd (channels’) dimension. This leads to a more local inverse problem, which is suitable for CNNs. We demonstrate our approach on sparse-view and limited-view CT, and show that it significantly improves reconstruction accuracy for any given network model. This allows achieving the same level of accuracy with significantly smaller models, and thus induces shorter training and inference times.
KW - ConvNets
KW - CT reconstruction
KW - Machine learning
UR - http://www.scopus.com/inward/record.url?scp=85151132951&partnerID=8YFLogxK
U2 - 10.1007/978-3-031-25066-8_29
DO - 10.1007/978-3-031-25066-8_29
M3 - ???researchoutput.researchoutputtypes.contributiontobookanthology.conference???
AN - SCOPUS:85151132951
SN - 9783031250651
T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
SP - 513
EP - 526
BT - Computer Vision – ECCV 2022 Workshops, Proceedings
A2 - Karlinsky, Leonid
A2 - Michaeli, Tomer
A2 - Nishino, Ko
Y2 - 23 October 2022 through 27 October 2022
ER -