Note:

The course presents the basic analytical and algorithmic tools used for processing information from a wide variety of physical sensors and applications ranging from medical CT scanners to speech signals. The basic theme of the course is the formulation of signal processing problems as inverse problems, and the solutions of inverse problems using the techniques of signal and system modeling along with parameter and signal estimation. The course also incorporates a number of computer-based laboratory exercises so that students can better understand how to implement the methods discussed in the class. The course is intended to be accessible to students with a variety of applications backgrounds, but they should have basic familiarity with probability, random variables, and random processes at the level of ECE 60000 or ECE 30200.

Counts as:

Normally Offered:

Each Fall

Campus/Online:

On-campus only

Catalog Description:

This class presents a collection of mathematical and statistical methods that form the foundation of modern computational imaging research and applications. Computational imaging seeks to form images from sensor data and is widely used in applications including consumer imaging, scientific imaging, industrial inspection, and security imaging. This class provides an advanced treatment of computational imaging based on an inverse-problems framework and blends perspectives from applied math, statistics, physics, and applications. The topics covered include stochastic modeling of images, Bayesian estimation, inverse methods, optimization, convexity, majorization techniques, constrained optimization and proximal methods, plug-and-play methods for advance prior models, the EM algorithm, Bayesian networks, Markov chains, hidden Markov models, and stochastic simulation. The underlying theory is presented in the context of applications including image restoration, tomographic reconstruction, clustering, classification, and segmentation.

Lecture Outline:

Assessment Method:

none