The capability of retrieving the image/signal of interest from extremely low photon flux is attractive in scientific, industrial, and medical imaging applications. Conventional imaging modalities and reconstruction algorithms rely on hundreds to thousands of photons per pixel (or per measurement) to ensure enough signal-to-noise (SNR) ratio for extracting the image/signal of interest. Unfortunately, the potential of radiation or photon damage prohibits high SNR measurements in dose-sensitive diagnosis scenarios. In addition, imaging systems utilizing inherently weak signals as contrast mechanism, such as X-ray scattering-based tomography, or attosecond pulse retrieval from the streaking trace, entail prolonged integration time to acquire hundreds of photons, thus rendering high SNR measurement impractical. This dissertation addresses the problem of imaging from limited photon budget when high SNR measurements are either prohibitive or impractical. A statistical image reconstruction framework based on the knowledge of the image-formation process and the noise model of the measurement system has been constructed and successfully demonstrated on two imaging platforms – photon-counting X-ray imaging, and attosecond pulse retrieval. For photon-counting X-ray imaging, the statistical image reconstruction framework achieves high-fidelity X-ray projection and tomographic image reconstruction from as low as 16 photons per pixel on average. The capability of our framework in modeling the reconstruction error opens the opportunity of designing the optimal strategies to distribute a fixed photon budget for region-of-interest (ROI) reconstruction, paving the way for radiation dose management in an imaging-specific task. For attosecond pulse retrieval, a learning-based framework has been incorporated into the statistical image reconstruction to retrieve the attosecond pulses from the noisy streaking traces. Quantitative study on the required signal-to-noise ratio for satisfactory pulse retrieval enabled by our framework provides a guideline to future attosecond streaking experiments. In addition, resolving the ambiguities in the streaking process due to the carrier envelop phase has also been demonstrated with our statistical reconstruction framework.
If this is your thesis or dissertation, and want to learn how to access it or for more information about readership statistics, contact us at STARS@ucf.edu.
Doctor of Philosophy (Ph.D.)
College of Optics and Photonics
Optics and Photonics
Optics and Photonics
Length of Campus-only Access
Doctoral Dissertation (Open Access)
Zhu, Zheyuan, "Computational Imaging with Limited Photon Budget" (2019). Electronic Theses and Dissertations, 2004-2019. 6864.