EECS 755 course policies F13

Topics in Signal Processing: Model-based image reconstruction methods

MW 10:30-Noon, 3433 EECS. (3 credits)

Instructor:	Professor Jeff Fessler,	 Email: fessler AT umich DOT edu
		Office Hours: (see web site) in 4431 EECS, Phone: 763-1434

Text: Draft of chapters of book: Image reconstruction algorithms and analysis,
supplemented by papers from the literature, presented in part by students.

Goal: Bridge gap between EECS 516,556,564 and modern image formation literature.

Topics:	Image restoration; tomographic image reconstruction (X-ray CT, PET);
	reconstruction from Fourier samples and MR image reconstruction;
	regularization; special optimization algorithms; analysis of
	spatial resolution / noise / detectability for nonlinear algorithms;
	compressed sensing and reconstruction from under-sampled data.
	Denoising, deblurring, super-resolution, and other inverse problems
	such as imaging time-varying objects (dynamic imaging).

Prerequisites: at least one of EECS 516, 556, or 564, or instructor permission.
Two of the three would be even better, and the ideal preparation would be
all three courses!

Enrollment / prerequisites / auditing.
Students who have had any of EECS 516, 556, 564 should enroll for grade; you
may audit the class only with instructor permission.  Students who have not
had any of those courses (or an equivalent) may audit if enrolled officially
as a "visitor," space permitting.

Grading:
	40% Homework (menu based)
	30% Project
	30% Participation
	Examples of participation (in case it is not obvious):
		attending class (when not traveling to conferences),
		asking questions in class and/or office hours,
		engaging in active learning activities in class,
		finding errors in book, suggesting problems,
		providing latex source for needed problem solutions,
		solving "open problems" mentioned in the book
			(some of which could merit a publication),
		implementing and evaluating or comparing algorithms in the book.

Collaboration policy and honor code: students are highly encouraged
to discuss homework problems and solutions with each other.  However,
in the end each student must turn in their own solutions (no copying).
Students are also highly encouraged to discuss project ideas and
approaches with each other.  However, unless prior arrangements are made
with me, students must do the "real work" for the project themselves
and turn in an individual written report (4-5 IEEE conf. paper style)
and make an individual presentation.

Curtis R Vogel, SIAM 2002
Computational methods for inverse problems

Heinz W Engl Martin Hanke Andreas Neubauer, Kluwer Dordrecht 1996
Regularization of inverse problems