CPSC 535/635 - Fall 2006

Outline/Reading:

Wiki:

Assignments:

General Instructions

• The first assignment is required and has a hard deadline (see below).
• You can select which of the remaining assignments you want to do.
• Your grade for the assignment portion of the course will be computed from the first assignment plus your best four grades from the others.
• After assignment 1 you may hand in assignments at any time but the following conditions apply
  • You may only hand in one assignment per week, and
  • All assignments are due before the end of term by the deadline indicated below.
  Note that this means that it would be unwise to wait until the last week of classes to hand in assignments because only one will be graded.
• Limit the written portion of any assignment to 2 pages, 12pt text, 1 inch margins. You may submit additional pages for code and plots. The goal is to force you to be succinct and save the marker's time.

Hand-in

Hand in assignments to your TA, Jim Young by email (jyoung at cpsc dot ucalgary dot ca), or check the course wiki for instructions.

This assignment is required:

• Assignment 1 - Introduction to Matlab and Octave (.ps or .pdf)

• Assignment 2 - The Discrete Fourier Transform (.ps or .pdf). Try this sample pgm image for part 2. You do not have to convert it from bmp.
• Assignment 3 - Inverse Filtering for Image Restoration (.ps or .pdf) - files
• Assignment 4 - Pattern Recognition I: Character Recognition (.ps or .pdf) - files
• Assignment 5 - 3D Computer Vision: Follow the Bouncing Ball (.ps or .pdf) - files
• Assignment 6 - Pattern Recognition II: Counting Beans (.ps or .pdf) - files
• Assignment 7 - Image Registration (.ps or .pdf) - files
  Mann and Picard - a paper describing one method to do registration automatically.
• Assignment 8 - Stereo Vision (.ps or .pdf) - files

Important Dates:

Miscellaneous Material:

• optimizatin and least-squares notes
• sample images
  • aaa.pgm
  • apple.pgm
  • books0.pgm
  • books1.pgm
  • dino_p.pgm
  • fallingwaterdistant.pgm
  • house.pgm
  • koa.pgm
  • left.pgm
  • lena.pgm
  • mona2.pgm
  • mona.pgm
  • nickle.pgm
  • p_3d.pgm
  • p_nsa.pgm
  • p_outer.pgm
  • p_sky.pgm
  • p_sw21.pgm
  • right.pgm
  • washington.pgm
  • washington2.pgm
  • zoo.pgm
• octave for windows - try Octave Forge for Windows - but beware, it may conflict with Cygwin.
• gnuplot tutorial
• Fast octave resampling using bi-linear interpolation (resample.m)
• 2D convolution in octave - this is almost compatible with the Matlab version - not necessary for octave 2.9 (conv2.m)
• Component labelling algorithm for octave. This is written as a C extension to octave because the algorithm does not vecorize well. Use mkoctfile to build the executable. This is almost compatible with the Matlab bwlabel. Distributed with octave 2.9. (bwlabel.cc)
• An easy-to-use image file format is pnm. You can get information about this file format on a Unix system using:
  • man pnm
  • man ppm
  • man pgm
  • man pbm
  I have written some octave function to read and write pnm/png files
  • pnmread.m - reads a pnm file.
  • pnmwrite.m - writes a pnm file.
  • pngread.m - reads png files - uses pnmread and netpbm utilities.
  If you are stuck using an older version of octave, try these
  • pnm_read.m
  • pnm_write.m
  • pnm_getint.m
  • pnm_getc.m
• In-class Fourier series demonstration - fourier1.m
• In-class convolution demonstration - convdemo.m
• In-class resampling demonstration - resample.m, resamplenn.m, resampledemo.m, resamplenndemo.m, rowcoords.m, colcoords.m, mona.tif
• Pixel matching demo.
• Examples of Lucas-Kanade least-squares optical flow and edge detection on a hockey sequence.
  • hockey.mpg - the original sequence
  • u.mpg - optical flow in the x direction
  • v.mpg - optical flow in the y direction
  • mask.mpg - validity of optical flow - black is no flow measurable, grey is gradient flow, and white is least-squares flow
  • edges.mpg - edge map for the original sequence - indicates large gradient values
• Karhunen-Loeve transform, covariance matrices, Mahalanobis distances example - kl.m and ellipse.m
• Shannon Jaeger wrote this modified version of image.m for use in octave along with some instructions.
• Documentation for an MPEG codec. Good chapter on DCT compression. In .ps too.

Links:

• Matlab and Octave
  • Mathworks, vendors of Matlab.
  • Octave
  • GNU Octave Repository: lots of good stuff for octave, helps with Matlab compatibility.
  • Peter Kovesi's MATLAB and Octave Functions for Computer Vision and Image Processing (includes Harris corner detector and lots of other useful stuff).
• Some Matlab Tutorials:
  • indiana.edu
  • math.utah.edu
  • sr.unh.edu
• CMU Robotics Computer Vision Home Page a good starting point for computer vision questions.
• an excellent phasor demo that can help visualize sums of sinusoids in Fourier analysis.
• A useful resource on The Kalman Filter
• Several on-line books are available that may be of interest.
  • D.H. Ballard, C.M. Brown; Computer Vision , Prentice-Hall Inc New Jersey , 1982 , ISBN 0-13-165316-4 .
    
    
  • D. Phillips; Image Processing in C: Analyzing and Enhancing Digital Images, RandD Publications, 1994.
    
    
  • D. Vernon; Machine Vision : Automated Visual Inspection and Robot Vision, Prentice Hall, 1991, ISBN 0-13-543398-3.
    
    
  • E.W. Weisstein; Mathworld , Online book.
    
    
  • I.T. Young, J.J. Gerbrands, L.J. van Vliet; Image Processing Fundamentals , Online book.
    
    
  • J. G. Moik, Digital processing of remotely sensed images, NASA, 1980.

Text: