|
|
Owen Carmichael : Teaching : ECS 289HAdvanced Image Processing and AnalysisUC Davis, Spring 2008 |
||||||||||||||||||||||||||||||||||||||||||||||||||||||
Home / NewsAffiliations / ContactResearchPapersTeachingCommunity ResourcesReading GroupsLeisure |
Lecture: MWF 4:10 – 5 PM Wellman 207 Office Hours: Tuesday 12-2 PM, 612E School of Medicine
Neurosciences Building Course web page:
Available from MyUCDavis Number of units: 4 Course overview
This
course gives an overview of state-of-the-art methods for automated extraction
of useful information from images.
ÒImagesÓ include data acquired from photographic cameras, 3D surface
range sensors, and volumetric biomedical imaging devices. The course material
emphasizes core principles and methods that are shared between imaging
modalities, and therefore is designed to be relevant to a broad range of
graduate students whose research involves imaging data: students interested
in computer vision, graphics, visualization, geometric modeling, medical
imaging, applied mathematics, and imaging-related biomedical sciences are all
encouraged to enroll. Students
will need to have a fairly strong grasp of linear algebra and basic calculus,
but no prior course material in imaging is strictly required to enroll. The
course material will be tuned to the interests of enrolled students, but it
tentatively covers a range of image processing tasks, including low-level
data pre-processing; mid-level grouping, alignment, and feature extraction;
and high-level object detection and retrieval. See below for a tentative course schedule. Key components of the course
This is essentially a
reading course in which each week is dedicated to a particular problem area
in image processing. Two papers that typify the current state of the art in
that area are featured each week.
The Monday and Wednesday lectures provide the background information
that students will need to understand the key attributes of the problem area
and the paper material. Each
Friday, one student will give a presentation that summarizes one of the two
assigned papers; the rest of the students read the other paper and write a
one-page summary of its contents.
I expect that the presentation will take about half of the Friday
lecture period and discussion of the other paper will take the other half. 1. Lectures: The Monday and Wednesday lectures are designed to give a broad
overview of the problem of the week, including its relevance, key issues or
aspects of the problem, variants of the problem, major schools of thought,
controversies, and so on. I will
also use this time to provide background that is more specifically needed to
understand the assigned papers.
In particular, difficult / unusual math and unclear exposition are two
aspects of the papers that I will try to deal with on Monday and Wednesday. 2. Paper summaries: Each week, all the students who are not making
the Friday presentation will read the paper that the presenter of the week
has decided not to
present. By the start of class
on Friday, all of those students are expected to email me a one-page summary
of the paper. The paper summary
should include these sections: a. The specific problem the paper addresses b. Key limitations to previous solutions c. The innovation of this paper: which of the prior
limitations does it overcome and how? d. The method: how does it work? e. Key experimental results f. Editorial: Was the paper convincing? Are there any glaring weaknesses or
questions that it should have addressed? Etc. This
format does not make much sense when we read a review paper; they donÕt
present one particular algorithm that you can describe, for example. Paper summaries for the review articles
can be more open-ended. The student making the presentation of the week
does not need to turn in a paper summary that week. In addition, students may skip one other paper summary of
their choosing at no penalty.
Since there are lecture 10 weeks in the quarter, this means each
student is required to write 8 paper summaries. On Friday, students should be ready to discuss
the paper they did the summary on.
Discussion of this paper will take place after the presentation of the
week. 3. Presentations:
Each student is expected to make one paper presentation; students will
sign up for a Friday presentation slot on a first-come, first-served basis.
The presentation of the week should consist mainly of slides shown on the
projector, along with written material on the board as needed. The presentation should cover the
same points as listed above for the written summaries. Presenters should plan on preparing
about 15-20 minutes worth of slide material; a good rough estimate is to have about one slide per minute,
i.e. roughly two or three slides each for the problem addressed, limitations
to previous solutions, etc. The presenter can choose to make the presentation
on his or her own laptop, or show it on mine. Presenters who want to use my
laptop should let me know beforehand if the slide format is something other
than PDF, PowerPoint, or Keynote; I need to make sure that my machine can
show the slides properly. 4. Final Project: A
final project, determined in consultation with me in the middle of the
quarter, will be required of each student. Each student is free to choose whether he or she wants to
write a 7-10 page paper or do a programming project. The paper is expected to be a
treatment of a particular problem area—either a subset of an area I
talked about in lecture or a completely separate one related to image
processing—that supplies the problem area with some sort of structure, e.g. a taxonomy of existing approaches, or a catalog
of key issues that must be dealt with.
The programming project should be an implementation and evaluation of
one or more image processing algorithms, along with a writeup of the
results. A key principle of the
final project is that if it is at all possible to design it so that it
advances your own research agenda in some way, you should absolutely do
so. Students who are involved in
a specific research project or lab, but are unsure how to design a course
project that might possibly dovetail with it, should talk to me about it. 5. Office hours: The lectures are there to get you up to speed on the paper
material, even if you have little background in image processing; still, some
of these papers might be difficult to understand anyway. Office hours are
held early in the week to give you a chance to get a better grasp of the
papers while you still have plenty of time to work on your summary or
presentation. Picking up the
paper for the first time on Thursday night is not an especially good idea. Sending me an email saying ÒI donÕt
understand the paperÓ on Friday after a week of complete silence on your part
is an even worse idea. Grading
Schedule
Students will fill
out a survey during the first week of the course that includes questions
about what specific image processing topics they are interested in learning
about. The tentative course
schedule, shown below, will be modified as needed to fit the indicated
interests. As is, the course
schedule is designed to move from Òdown to up;Ó that is, we start with
low-level removal of noise from images, move into the extraction of coherent
regions, ÒinterestingÓ features, and alignment of multiple data sets, and end
up talking about the extraction of higher-level, more
semanatically-meaningful information such as the locations or categories of
objects in the image. PDFs of
all papers will be available on the password-protected course website on
MyUCDavis; please do not distribute the PDFs to anyone else. I will also put PDFs of the Monday
and Wednesday slides up on the MyUCDavis site; please donÕt distribute those
around either.
Academic Integrity
Each student must produce
their own individual paper summaries, final project, and presentation. Discussing these assignments with
classmates is fine but outright copying from other students or any other
source is forbidden. Any
instance of suspected cheating or plagiarism will be referred to the Office
of Student Judicial Affairs for adjudication. The "Code of Academic
Conduct" describes relevant policies and procedures. (A copy of this
document can be obtained through the Office of Student Judicial Affairs, 752-1128;
http://sja.ucdavis.edu.). |
