EE520: Special Topics in Comm. & Signal Processing, Fall 2005

(Signal Processing for Image Analysis and Computer Vision)

The course is about answering the following question: Given an image, a set of images or a time sequence of images (video) what can you infer about the scene?  "Inference" includes both detection problems such as object recognition (or more generally pattern recognition) or image retrieval, and estimation problems in both 2D and 3D.  2D estimation includes segmentation, registration and tracking in 2D, while 3D estimation includes reconstruction of the 3D scene, which may be static (stereo) or moving (structure from motion). In addition we will also spend some time on shape analysis, shape matching and shape tracking. "Image Analysis" is a generic term which refers to all of the above, while "Computer Vision" usually refers to deducing the properties of the 3D world from one or multiple images.


Prerequisites - Basic signal processing, probability/statistics, calculus. Estimation and Detection theory may be useful

Instructor - Dr. Namrata Vaswani,   Email: namrata AT,   Phone: 515-294-4012,   Office: 3121 Coover Hall


Class Time and Room: Monday-Wednesday,12pm - 1:20pm in 1207 Coover Hall

Office Hours: 3121 Coover Hall, Mondays, 2 - 3:30pm or any other time b/w 2-7pm on Mondays and Wednesdays (call to make sure I'm in) or any other day.

Discussion Session: 5:15 - 6pm Mondays in Sproul (2202 Coover):  Plan to discuss (i) questions related to what's taught in class, (ii) MATLAB or general implementation issues not addressed in class,  (iii) problems from your research, or (iv) projects. Those auditing are also encouraged to come and participate.

If anyone shows up, please call me or come to my office.



Midterm Exam: October 17 (Monday)

Topics: Edge detection, Calculus of Variations, Optical Flow, Segmentation, Registration, Least Squares, Tracking

Everything that is on the handouts and sections of books/papers that any handout asks you to read


Project Deadlines, Details  Please start deciding your projects!


Syllabus, Evaluation

Main Reference Books,      Books on Reserve in the Library



Homework 1    

Homework 2  

Homework 3

(NEW) Homework 4


Handouts: (These may have errors/typos, please please email me if you find any)

Introduction (ppt) 

Edge Detection 

Calculus of Variations  

Optical Flow  

Kalman and Extended Kalman Filter   

Particle Filter

Least Squares




Some Scanned Notes: (read at your own risk! Alternate pages are inverted, print carefully)

Segmentation: Parametric Active Contours, Geometric Active Contours, Level Set MethodDifferent types of methods

Registration:  Landmark ShapeLeast Squares and Kalman filtering intuition



Reading Material:  These are only very small subset of the many papers or books on any topic.


Edge Detection -  Chapter 9 of A.K. Jain's book, Section 4.3 of Sonka et al's book

Texture - Section 14.1 of Sonka et al,  Manjunath, & Chellappa's paper

Optical Flow - Chapter 5 of Tekalp's book,  Chapter 14 of Sonka et al's book, Weblink


Segmentation & Registration
    Sections 5.1-5.3,  Section 8.2 of Sonka et al's book, Chapter 1 of Sapiro's book,

    Weblink 1,    Weblink 2 

    Kass,Witkin,Terzopolous, IJCV 1987  (Parametric)

    Kichenassamy et al ICCV 1995  ("Geometric" Edge-based)

   Yezzi, Tsai, Wilsky, LIDS Technical Report, 1999    ("Geometric" Region-based)

    Chan,Vese, Trans IP 2001  ("Geometric" Region-based)

    Yezzi,Zollei,Kapur 2001Yezzi,Soatto, IJCV 2003  (Joint Registration & Segmentation)

    Level Set Method:

        Sethian's Introduction page 

        Publications page (specifically see the Narrowband Level Set Methods and Extension Velocities papers)

Statistical Shape Analysis: Dryden & Mardia's book, Chapter 2, 2.1-2.2, Chapter 3, 3.1-3.4, Chapter 4, 4.1, 4.3


Kalman & Extended Kalman Filter:

    Welch & Bishop (easy to read)

    Contour tracking application

    EKF for SfM - Broida & Chellappa

    Kalman's first paper

    Multiple Hypothesis Trackers

Particle Filters:

    Doucet 1998 Technical Report,  

    Tutorial paper

    First few papers: Gordon et al 1993Kittagawa 1996,  

    Condensation (first computer vision application), webpage

    Sequential Monte Carlo page   Sequential Monte Carlo Book

3D Reconstruction:

Camera Models:                                             Hartley & Zisserman, Chapter 6, 6.1 - 6.4
Epipolar Geometry:                                        Hartley & Zisserman, Chapter 9, 9.1 - 9.5
Reconstruction of Cameras & Structure:      Hartley & Zisserman, Chapter 10, 10.1 - 10.3
Estimating the Fundamental Matrix:             Hartley & Zisserman, Chapter 11.1 - 11.4.1

Structure  Reconstruction:                             Hartley & Zisserman, Chapter 12.1 - 12.4


Old talk with too much detail: Particle Filtering and Change Detection



Other Related Course pages  (with lots of useful handouts and links):

    EE527 (Detection and Estimation Theory)

    STAT580 (Computational Methods on Statistics)   - see for handout on Monte Carlo methods

    EE524 (Digital Signal Processing)

    EE523 (Random Processes for Communications and Signal Processing)