EECS 556	Image Processing
	Winter 2005	Room 1005 Dow, MWF 9:30-10:30AM

Instructor:	Professor Jeff Fessler
		Email: fessler AT umich DOT edu
		Office: 4240 EECS
		Office Hours: (see web site)

Web:		http://www.eecs.umich.edu/~fessler/course/556
		handouts, solutions, errata, m-file templates, past exams, etc.

Text:		Lim: 2D signal and image processing, Prentice Hall, 1990.
		This is optional.  Other references on reserve at library.

Objective:	Fundamentals of imaging and image processing

Topics:		Image formation, sampling, modeling/representation,
		enhancement, restoration, analysis, and compression.

Prereq.:	EECS 551 (DSP/linear algebra), EECS 501 (random processes)

Grading:	Homework / Computer Problems	25%	(see policies below)
		Exam 1				25%	mid Feb.
		Exam 2				25%	mid Mar.
		Project				25%
		In all of the above, legibility counts.
		Scores may be standardized before computing the final score
		if the means and standard deviations vary.
		The exams may be a combination of in-class and open book
		take-home, and will include Matlab problems.

	Requests for re-grades of exams must be submitted in writing within
	one week of exam return.  All questions may be re-graded.
	Letter grades will be assigned using a curve.
	The cutoff for an A- will be 90% or lower.

Project: In liew of a final exam, students will work in small groups on
	image processing projects that apply the tools learned in the
	course as well as using ideas from the contemporary literature.
	A written proposal for the project will be due part way through
	the semester.  A written report will be due on Apr. 25.
	Student groups will give oral presentations to the class Apr. 15-20.

Email:	Announcements will be mailed to eecs556@eecs.umich.edu and you must
	send email to eecs556-request@eecs.umich.edu with the word "subscribe"
	in the subject line to receive such announcements.



Notes:

Some subset (possibly all) of the problems from each assignment will be graded.
Solutions will be provided for all problems.

ABSOLUTELY NO LATE HOMEWORK ASSIGNMENTS WILL BE ACCEPTED.
However, I will drop your lowest homework score when computing grades,
which gives some flexibility if you become ill etc.

Homework/computer project policy.  You must attempt to solve all
homework problems by yourself, and implement all computer programs
by yourself.  Copying homework solutions from another student or from
solutions from previous semesters will be considered violations of the
engineering honor code.  However, after making a genuine attempt to
solve the homework problems, you are encouraged to discuss the answers
with other students currently enrolled in 556 to check the answers and
compare solution approaches.  After such a discussion, you may rewrite
your answer as long as you do so individually, without referring to
the solutions of other students or to solutions from previous terms.
Basically, the answers you turn in should reflect your own level of
understanding, not someone elses.  All solutions submitted must be
generated by the person whose name appears on the assignment.

Homework grading.  The GEO contract requires that a GSI title be given to
any graduate student who grades papers "in a manner that requires subjective
evaluation above and beyond the mechanical or routine comparison of submitted
papers or examination with answers, responses, or elements predetermined as
correct or acceptable."  Past use of graders has led to grievances.  To comply
with this GEO contract, in this course HW scores for a N-point problem will
be assigned by the grader as follows: 0 points if no work shown, 1 point if
some work is shown that is relevant but largely incomplete or incorrect,
N-1 points if correct except for a very minor mistake such as a minus sign,
N points if completely correct.

All students must take all exams during the scheduled times.

No course incompletes will be given, except per UM regulations.

Textbooks available at library:
bracewell:78 bracewell:95 goodman:68 jain:89 lim:90

What sections are covered?  On the web page I will frequently update
a lecture-by-lecture list of what topics I cover each lecture.

At least half of the homework assignment points will be Matlab
related "projects."  And the points assigned to some of these problems
may far outweigh those of other such problems. 

The treatment of image coding may be brief.  Students with primary interests
in compression should take EECS 651 (source coding) and whatever class
eventually replaces the former version of EECS 551 (wavelets), since JPEG 2000
is a wavelet-based coding method.  Since neither wavelets nor 651 is a
prerequisite for this course, a detailed treatment of the JPEG 2000 standard
is infeasible, but we will cover the general ideas of transform coding.