Computational imaging consists of joint design of measurement strategy and estimation algorithms for image formation from radiation fields. This course reviews principles of forward model and inversion algorithms for computational imaging and analyzes imaging systems for geometric, wave and statistical radiation field models. Forward models, consisting of discrete representations of continuous image and measurement spaces, are fundamental to computational imaging. The course reviews how to form and evaluate such models. Since convolutional neural networks are the most important tool in modern inverse models, their use and application in concert with linear and regularized regression is explored. Coded aperture and structured illumination systems are considered for X-ray imaging and coherent phase retrieval. Physical compression and decompressive estimation are discussed. The physical and information limits of imager resolution are explored.
This course is offered Fall semesters. Computer Code used in the course is available at
https://github.com/arizonacameraLab/computationalImaging
Computational photography blends computer algorithms with traditional photography to create images that are not
feasible with traditional digital imaging. This course examines spatial, temporal and spectral sampling strategies in photography and video. Dynamic control, multiframe estimation and array camera design are discussed. Plug and play neural processing, video transformer networks, reinforcement learning and other neural image processing techniques are explored.
This course is offered in Spring Semesters.