Intructor: Dr Sohaib A. Khan
sohaib at lums dot edu dot pk
http://web.lums.edu.pk/~sohaib


TA:
Salman Cheema
salmanc at lums dot edu dot pk

Kashif Javed
kashif@lums.edu.pk

Lecture 1 [download]:

• Course introduction and policies
• Introduction of Computer Vision area
• The complexity of high-level vision tasks
• Human perception, and cues that aid in perception
  
• Modules of the course

Lecture 2 [download]:

• Image Representation
  Image coordinates
  Homogeneous coordinates
  Binary, Gray-scale and Color images
  Example operations on images
• Image Histogram (gray-scale histogram, color histogram)
• Cameras
  Aperture
  Shutter speed
  Digital cameras
  Video cameras
  Interlacing, progressive scan
  Human eye
• Portable Bit Map Formats
  PGM, PPM file formats
  Reading/Writing binary pgm/ppm files

Assignment 0 (no credit)
Write functions ReadBinPGM, ReadBinPPM, WriteBinPGM, WriteBinPPM in C/C++

Example PPM/PGM Files: mecca06.pgm, mecca06.ppm

Light experiment files (PPM):
right-lamp, left-lamp, ambient

Irfanview is a good image viewer. You can
get it from http://www.irfanview.com/

Format specification:
[pbm.txt][pgm.txt][ppm.txt]
[Thanks to Numan Sheikh for sending me these. See www.wotsit.org for these and more format specs]

Reading
Shapiro Ch 1, 2.1-2.6
UCF Vision lab projects page

Lecture 3 [download]:

• Transformations
  Example Application: Image Registration
• Transformations in 2D
  Scaling
  Rotation
  Translation
  Shear in x- and y-direction
  Inverse transformations
• Cocatenation of transformations
  Rotation about arbitrary point
  Importance of order

Lecture 4 [download]:

• Affine Transformation
• Rigid Body Transformation
• Recovering the best affine transformation from correspondences
  Least Squared Error solution
  Pseudo Inverse
• Projective Transformation
• Other displacement models
  Biquadratic
  Bilinear
  Pseudo Perspective

Reading: Shapiro 11-11.4

Steve Mann's paper on video orbits [web][local] contains a nice introduction to these transformations (see sections I and II of this paper)

Lecture 5 [download]:

• Displacement models
• Warping
  Bilinear Interpolation
• 3D Translation and Scaling
• 3D Rotation
  Rotations about principal axes
  Small angle approximation
  Rotation about arbitrary axis
  Rodrigues' Formula

Homework 1: Image Warping [handout]
Due: Wed Sept 22, 2004 at 5:00

Test images
[mecca06.pgm][mecca06t.pgm][mecca06.ppm]

Reading: Shah 1.1-1.3, 1.8

Lecture 6 [download]:

• Rodrigues' Formula
• Eigenvalues and Eigenvectors of Rotation Matrix

Lecture 7 [download]:

• Frames of Reference
• Pinhole Camera
• Perspective Transform
• Camera Model

Reading: Shah 1.4-1.5, Gonzales handout

Quiz1 [download]

Lecture 8 [download]:

• Camera Models
• Orthographic Projection
• Interpretation of Projective and Affine Projections

Lecture 9 [download]

• Camera Calibration
• Stereo (introduction)

Lecture 10 [download]

• Stereo
  Concept of Disparity
  Epipolar Constraint
  Basic Stereo Derivations
  Results
• Binary Image Processing
• Applying a mask on an image
• Average Filter
• Edge Detection
  

Lecture 11 [download]

• Filtering
  Average Filter
  Gaussian Filter
  Removing Noise
• Differentiation
  Discrete approximation
  Gradient Vector
• Comparison of filtering and differentiation masks
• Sobel, Prewitt, Robert's operators
• Detection of Edges
• Gradient Magnitude and Direction
  Interpretation
  

Homework 2 : Problems, Module 1 [handout]
Due: Wed Oct 6, 2004 at 5:00 pm

Important Annoucement: Midterm on Sat Oct 9 (out of class). Please mark your calendars.

Lecture 12 [download]

• Properties of Ideal Edge Detector
• Canny's Edge Detector

Lecture 13 [download]

• Marr-Hildreth edge detector (Laplacian of Gaussian)
• Finding lines in binary images
  Least-Squared Error solution and its problems
• Hough Transform for finding lines
  Concept of parameter space (dual)
  Algorithm using (m,c) parametrization
  Algorithm using (p, theta) parametrization

Lecture 14 [download]

• Hough transform for finding lines
  (p, theta) parametrization
• Hough transform for detecting circles, and other parametric shapes
• Centroid and Area
• Generalized Hough Transform
  Constructing the R-Table
  Detecting shapes based on R-Table

Lecture 15

• Discussion of homework solution

Quiz 2 [download]

Midterm Exam tomorrow (9th Oct)
8:45-10:00 am, in A-11
Closed book / closed notes / calculator allowed

Lecture 16 [download]

• Rotation and scaling invariance for GHT
• Binary Morphology
  Dilation
  Erosion
  Opening
  Closing
  Logical Operations: AND, OR, Subtraction
• Example application using morphology
• Connected Component Labeling
  Recursive code
  Iterative code
• Region Properties
  Area
  Centroid
  Perimeter
  Length of Perimeter
  Circularity

Homework 3: Canny Edge Detector [handout]
Due: Fri Oct 22, 2004 at 5:00 pm

[Shapiro] has a good description of binary operations in Ch 3. The 'gears' example is also taken from [Shapiro]. Region properties and connected component algorithms are also discussed in Ch 3 of [Shah].

Lecture 17 [download]

• Region Properties
  Circularity
  2nd Moments
• Motion
  Brightness Constancy Assumption
  Brightness Constancy Equation
  Interpretations of BCE
  Aperture Problem
• Optical Flow
• Lucas-Kanade Method for computing optical flow

Reading: Motion Handout

Lecture 18 [download]

• Limitations of BCE
• Pyramids
  Reduce Operation
  Expand Operation
  Aliasing
• Lucas-Kanade with Pyramids

Reading: Motion Handout

Lecture 19 [download]

• Global Flow
• Affine Global Flow (Bergen/Anandan/Hanna/Hingorani's Method)
  Iterative Refinement
  Blending
• Application
  Video Mosaics
  Stablization,
  Hierarchical Dominant Motion Estimation

Reading: Motion Handout

Lecture 20 [download]

• Recognizing Facial Expressions (Black and Yacoob 1997)
• Projective Global Flow (Szeleski's Method)

Reference Readings: (you may go through these if interested)
Black and Yacoob's paper [web][local]
Szeleski's paper [web][local]

Lecture 21 [download]

• Discussion of Midterm
• Mann-Picard method for Projective Global Flow

Lecture 22 [download]

• Pattern Classification Examples
• Pattern matching using correlation
• The need for normalization term

Lecture 23 [download]

• Correlation in Frequency Domain
• Various Correlation Methods
• Distance Transform
• Medial Axis Transform
• Distance Measures
  • Euclidean
  • City Block
  • Chess Board
• Properties of a Distance Metric
• Hausdorff Distance
• Introduction to Bayes and Nearest Neighbor Classifiers

Readings: Material on classifiers is included in [Shapiro] Ch 4 and [Snyder] Ch 15. Distance transform and Medial Axis transform are described in [Shah] Ch 4

Homework 4 [download]
Due 10th Nov before class (9:50 am)

Dataset for Problem 6:
yosemite1r.pgm yosemite2r.pgm

Dataset for Problem 7:
LUMS sequence: as frames, as zip file

If using MATLAB for pr6&7, the following functions will be useful for reading images:
readbinpgm.m readbinppm.m

Lecture 24 [download]

• Bayes Classifiers
  • Example, human trackin
• Change Detection
  • Differencing consecutive frames
  • Background subtraction using single frame
  • Background subtraction using Gaussian model

Lecture 25 [download]

• Gaussian Background Subtraction
• Grimson's Method

Lecture 26 [download]

• Gaussian Background Subtraction
• Face Recognition (using PCA)

Lecture 27 [download]

• Face Recognition

Lecture 28 [download]

• Clustering
• KMeans
• Agglomerative
• Divisive
• Choosing K
• Segmentation based on color
• Segmentation based on motion

Homework 4 due