ECE497PM
Statistical Image and Video Processing
Fall 1998
11:30-1pm, Tuesday & Thursday, 170 Everitt Lab
Please check this site frequently
for announcements.
People
-
Instructor:
-
Pierre Moulin,
-
2265 Beckman Institute, Tel: 4-8366, email:
moulin@ifp.uiuc.edu.
-
Office hours: 10-6 Mon, Wed, Fri, send me email first
-
Teaching Assistant:
-
Prakash Ishwar,
-
email: ishwar@ifp.uiuc.edu.
Information
-
Homeworks.
-
Lecture Notes.
-
Bibliography.
-
Projects.
Announcements:
-
12/11: Project reports due Thu 12/17
at 8am Notes on Video Processing are available from EL
-
11/14: HW5 solution set is at TIS
HW6 is available above Notes on Image Restoration are available from EL
-
11/6: Midterm solution set is at TIS
Updated bibliography on Image Restoration
-
10/31: HW5 is available, due 11/10
Check new link to Projects above
Reminder: Midterm will take place on Tuesday 11/3, 7:30-9:30 PM in Room 170 EL
-
10/21: Full HW4 is available
-
10/16: HW3 solution set is at TIS
-
10/3: Final version of HW3 now available on Homeworks page
-
9/24: Solutions to computer experiments
now available on Homeworks page
-
9/22: First installment of HW3
available, due 10/6
.......Updated Useful Links
-
9/15: Final version of HW2 available
-
9/13: First installment of HW2 available
-
9/8: Prepare this Thursday's discussion
of image models, see Lecture Notes above
-
9/3: Change to HW1!
-
8/29: HW1 available, due 9/8
Course Outline
-
Multidimensional Signal Processing ( ~ 1.5 lectures)
Discrete and continuous Fourier transforms, Multidimensional (MD) random
processes, MD systems
-
Review of Human Visual Perception ( ~ 0.5 lecture)
Modulation Transfer Function, Visual Masking, Noise Visibility, Color
Vision; Distortion measures
-
Image Modeling ( ~ 5 lectures)
Continuous- and discrete-space models, including AR, MRF,
and nonstationary models; Karhunen-Loeve expansions; Transform-based models;
Edge and texture models; Doubly stochastic processes,
priors and hyperpriors, "universal" models
-
Image Sensor Models ( ~ 2 lectures)
Photographic film; electronic imaging; solid-state imaging systems,
Poisson/Gaussian noise models;
Mass-market and scientific CCD imaging applications;
Smart sensors; Coherent imaging applications, speckle noise models
-
Image Displays ( ~ 2 lectures)
Halftoning; CRTs and Flat-Panel Displays
-
Image Restoration ( ~ 6 lectures)
Wiener filtering technique and its limitations; Maximum-likelihood and
Bayesian estimators, basic properties;
Statistical ill-posed problems, regularization;
Model selection criteria, MDL principle;
Restoration of film, CCD images, and speckle images;
Blind deconvolution
-
Image Compression ( ~ 5 lectures)
Basics of Source Coding theory (lossless and lossy); Vector Quantization,
codebook design; Transform and Subband Coding, adapted bases; universal
coding
-
Video Modeling ( ~ 2 lectures)
3-D and 2-D Motion models; ill--posedness of inverse problem; block
matching (simple, hierarchical, and overlapped), optical flow;
Transform-based models, Motion-compensated prediction models
-
Video Restoration ( ~ 1 lecture)
motion-based techniques
-
Video Compression ( ~ 2 lectures)
Transform and motion-based techniques;
Bidirectional prediction; Adaptive coding techniques
Prerequisites: ECE434, ECE371PM or ECE447, experience with Matlab.
Recommended: ECE461, ECE451
Grading Policy
The final grade will be determined based on regular homeworks, one midterm
exam, and a final project:
-
Homeworks: 20%
-
Midterm Exam: 25%
-
Final Project: 55%
Homework Policy
You may use any material you want to solve homework problems. Teaming up
with other students to work out homework problems is acceptable. However,
you will maximize your learning experience by working out homework problems
on your own. In any event, you are required to write out your solutions
independently.
Homeworks will typically be due on Tuesdays. The TA will xerox an unspecified
number of HWs and grade them. Homework solutions will be posted at the
TIS copy store on Friday. Your (unmarked) homeworks will be returned to
you on Thursday. You will then grade your own homework and email the grade
to the TA by the following Friday at 12 noon. If there is a significant
discrepancy between your grade and the grade determined by the TA, a revised
grade will be emailed back to you, with copy to the instructor.
Resources
No textbook is required, but the books by Jain (Fundamentals of Digital
Image Processing) and Tekalp (Digital Video Processing) are recommended.
Articles in recent scientific and technical literature will also be used
as references. Additional class notes and homework solutions will be available
at the TIS copy shop.
-
Supplementary Materials:
-
A. K. Jain , Fundamentals of Digital Image Processing , Prentice-Hall,
1989.
-
A. M. Tekalp, Digital Video Processing , Prentice-Hall, 1995.
-
P. Moulin, ECE371PM Lecture Notes, Spring 1998.
-
H. Stark and J. W. Woods, Probability, Random Processes, and Estimation
Theory for Engineers, Prentice-Hall, 1994.
-
A. M. Yaglom, Correlation Theory of Stationary and Related Random
Functions I: Basic Results, Springer-Verlag, 1987.
-
L. Breiman, Probability, SIAM, 1992.
-
U. Grenander, Abstract Inference, Wiley, 1981.
-
H. V. Poor, An Introduction to Signal Detection and Estimation,
2nd Ed., Springer-Verlag, 1994.
-
A. Gersho and R. M. Gray, Vector Quantization and Signal Compression,
Kluwer, 1992.
-
M. Vetterli and J. Kovacevic, Wavelets and Subband Coding, Prentice-Hall,
1995.
Some of these books are also available at Reserve desk in the Grainger Library.
Useful Links