GPU acceleration for spherical grid tomography

Tomography reconstructs cross-sectional images from multi-directional data and plays a crucial role in fields such as medical imaging and solar physics. In solar physics, tomography is crucial for reconstructing 3-D electron density volumes of the solar corona from 2-D coronagraph images. This thesis investigates how factors like grid resolution, the number of viewpoints, and view angles influence reconstruction accuracy and computational cost in solar tomography. Through a series of controlled experiments using synthetic data, we explore the tradeoffs between these factors to provide insights into optimizing future solar tomography tasks. To enable these computationally intensive experiments, we develop a GPU-accelerated forward model framework specifically designed for spherical grids. This framework achieves a significant speedup of 27x to 270x compared to sequential implementations, allowing us to conduct large-scale simulations and reconstructions efficiently. This work not only advances the understanding of the solar tomography reconstruction but also paves the way for future advancements in space tasks.

Name thesis:degree_name

M.S.

Level thesis:degree_level

Thesis

Discipline thesis:degree_discipline

Electrical & Computer Engr

Grantor

University of Illinois at Urbana-Champaign

Year dc:date

2024

Author dc:creator

• Zhang, Zheyuan

Contributors dc:contributor

• Lumetta, Steven Sam
• Butala, Mark

• Gpu Acceleration
• Solar Tomography

Statement dc:rights

• Copyright 2024 Zheyuan Zhang

Language dc:language

en, eng

Handle dc:identifier

https://hdl.handle.net/2142/127161

OAI record oai:identifier

https://www.ideals.illinois.edu/oai-pmh?verb=GetRecord&metadataPrefix=etdms&identifier=oai%3Awww.ideals.illinois.edu%3A2142%2F127161