@phdthesis{oai:oist.repo.nii.ac.jp:00000186,
 author = {マームッド, ファイザル and Mahmood, Faisal},
 month = {2017-12-21, 2018-04-09},
 note = {Molecular structure determination is important for understanding functionalities and dynamics of macromolecules, such as proteins and nucleic acids. Cryo-electron tomography (Cryo-ET) is a technique that can be used to determine structures of individual macromolecules, thus providing snapshots of their native conformations. Such 3D reconstructions encounter several types of imperfections due to missing, corrupted and low-contrast data. This thesis focuses on the algorithmic and architectural aspects of improving and accelerating tomographic reconstructions specifically for Cryo-ET. The thesis explores modern compressed sensing and graph-based non-local approaches for noise removal and for partially recovering the missing wedge. These methods act as a proof-of-concept for the applicability of sparsity exploiting methods to tomographic image reconstruction. The thesis also explores, analyses and explains an extended field (EF)-based noise removal method. When used in conjunction with a variety of reconstruction procedures with a regularization capability it proved to be computationally efficient, reliable and stable. Through extensive empirical simulations it was shown that extending the reconstruction space reduces the error at a relatively lower regularization parameter thus allowing a better fit with the projections and preventing oversmoothing. Computational constraints are a major issue in speeding up tomographic reconstruction and refinement. One of the fundamental components, which often becomes a bottleneck in a variety of analytical tomographic reconstruction methods, is the 2D fast Fourier transform (FFT). Generally, 2D FFTs suffer from edge artifacts or series termination errors, which stem from the fact that two opposing edges of an image are often not periodic. These artifacts can propagate to next stages of processing and appear as errors in reconstructions. This thesis also explores simultaneous 2D FFTs and edge artifact removal for real-time applications. This was accomplished on a multi-FPGA (Field Programmable Gate Array) reconfigurable computing system with a high-speed bus. The algorithmic optimization and architecture are general and can be replicated t a variety of different hardware setups.},
 school = {Okinawa Institute of Science and Technology Graduate University},
 title = {クライオ電子トモグラフィーのためのアルゴリズムおよびハードウェアの開発},
 year = {}
}