막수 포럼 초청 세미나
일시: 2013년 11월 6일 (수), 12:00~13:00
장소: 전산학과(E3-1) 4층 오상수 영상강의실
Accelerated optimization algorithms for statistical 3D X-ray CT image reconstruction
Presenter: Kim, Donghwan (Electrical Engineering: Systems, University of Michigan - Ann Arbor)
X-ray computed tomography (CT) has been widely celebrated for its ability visualizing the anatomical information of patients, but has been criticized for its radiation exposure to the patients. Statistical image reconstruction can provide good images from low-dose scans, but requires very long computation time for minimizing an X-ray CT statistical cost function. Therefore, our goal is to develop fast optimization algorithms for statistical X-ray CT problems. One widely used optimization algorithm in tomography problems is an ordered subsets (OS) method that uses only a subset of a measurement data per iteration for computational efficiency. In particular, OS methods based on separable quadratic surrogates (OS-SQS) that are massively parallelizable are preferable considering the modern computing architectures, but it requires many iterations for the convergence. In this talk, I will present novel accelerated OS-SQS algorithms exploiting optimization transfer techniques and Nesterov's momentum algorithms. A real 3D helical CT scan is used to examine the acceleration of the proposed algorithms.