Title: Efficient Computing of Three-Dimensional Quantitative Phase Imaging

Committee:

Prof. Thomas K. Gaylord, ECE

Prof. Shyh-Chiang Shen, ECE

Prof. Christopher J. Rozell, ECE

Abstract: Quantitative Phase Imaging (QPI) is a powerful imaging technique for measuring the refractive index distribution of transparent objects such as biological cells and optical fibers. The quantitative, non-invasive approach of QPI provides preeminent advantages in biomedical applications and the characterization of optical fibers. Tomographic Deconvolution Phase Microscopy (TDPM) is a promising 3D QPI method that combines diffraction tomography, deconvolution, and through-focal scanning with object rotation to achieve isotropic spatial resolution. However, due to the large data size, 3D TDPM has a drawback in that it requires extensive computation power and time. In order to overcome this shortcoming, CPU/GPU parallel computing and application-specific embedded systems can be utilized. In this research, OpenMP Tasking and CUDA Streaming with Unified Memory (TSUM) is proposed to speed up the tomographic angle computations in 3D TDPM. TSUM leverages CPU multithreading and GPU computing on a System on a Chip (SoC) with unified memory. Unified memory eliminates data transfer between CPU and GPU memories, which is a major bottleneck in GPU computing. This research presents a speedup of 3D TDPM with TSUM for a large dataset and demonstrates the potential of TSUM in realizing real-time 3D TDPM.