1

CS 436Computer Vision Fundamentals

Sohaib A KhanRoom 215, sohaib@lums.edu.pk

Office Hrs: M 10am-11amW 10am-11:amTh 3pm-4pm

Course Status

Elective for Senior/Graduate studentsPrerequisites:

Strong programming backgroundMathematics Background

RequiredMatrix Manipulation, Calculus

Will be helpful, but not assumedProbability, Random Processes, Signal Processing…

2

Goals

Basic Goal: Generate excitement about CV

Introductory theory & applications of CVDemonstration of simple exciting applicationsAbility to write programs to solve CV problemsExamples of CV SystemsIntro to some current research topics

Text

Computer VisionLinda G. Shapiro, George C. Stockman ISBN 0-13-030796-3 Prentice Hall, 2001Additional course notes and reading material will available through the course website.Programming Environment:

C/C++ and MATLAB

3

Lectures

20 sessions, 100 minutes each1 in-class midterm, 1 final exam50 minutes each week for programming aspects of the courseNo grade for attendance

Grading

40% projects and programming assignments10% assignments/homework5% reading assignments

Graded through class participation/quizzes

20% mid-term25% final exam (non-comprehensive)

4

Cheating/Plagiarism

Will not be tolerated…

Course Website

Tool to keep all reading links, additional information in one placeLecture notes and slides will be availablehttp://web.lums.edu.pk/~sohaib/cvf-fall02.htm

5

What is Computer Vision

“The goal of Computer Vision is to make useful decisions about real physical objects and scenes based on sensed images”

ImageProcessing

Image IN Image OUT Computer Vision

Image IN

Symbolic Decision OUT

Computer Vision Area

Integration of several areasImage processingStatistical inferenceMachine intelligenceDecision theory…

Teaching methodologyCover some basic areasBreadth vs DepthShow lots of examples

6

Module 1 (4 Sessions)

Introduction (1)Transformations and Imaging Geometry (3)

2D Transformations

Module 1 (4 Sessions)

Global Transformation ModelsAffine, Projective, Bilinear

From http://wearcam.org (Professor Steve Mann)

7

Module 1 (4 Sessions)

3D Transformations

Camera ModelsCamera Calibration

Module 1 (4 Sessions)

WarpingApplying transformation to an imageCan be done for multiple images

8

Module 2 (6 Sessions)

Basic Binary OperationsThresholding, Morphology, Region Properties, Moments, Connected Component Labeling

9

Module 2Edge Detection

Shape Representation, Hough Transform

Module 3 (6 Sessions)

Pattern Recognition, Decision TheoryClustering

K-MeansBayes ClassifiersMAP EstimationEM AlgorithmPrinciple Component AnalysisFace Recognition

10

Module 3

Background Subtraction

Segmentation

11

12

Module 4 (4 Sessions)

MotionBrightness Constancy EquationOptical Flow

Lucas-Kanade MethodHorn and Shunck Method

13

Module 4

Global Motion Estimation / Image Stablization

Affine Least-Squares Method

14

Programming Assignments1. Image Warping2. Edge Detection3. Hough Transform for Lines4. Clustering5. Face Recognition6. Segmentation7. Lucas-Kanade Optical Flow8. Global Motion Estimation & Image Stabilization

All assignments may not be compulsory, depending on the ability of the students

Outline.doc

15

Reading Assignment 1

Chapter 1Section 2.5http://www.cs.ucf.edu/~vision/projects/projects.htmlhttp://wearcam.orghttp://wearcam.org/research.html

Digital Images

Images are essentially large matricesGray scale images are 2D matricesColor images have 3 layers (R, G, B)Each cell of the matrix represents a pixelEach pixel is quantized to a set of values, e.g. from 0-255 (1 Byte per pixel)

16

PBM, PGM, PPM Format

Portable Bit Map FormatsSimple Image header in ASCIIImage data in either ASCII or Binary.PGM for Grayscale images.PPM for Color (RGB) images

Programming Assignment 0

Write Program to Read/Write BINARY .ppm & .pgm filesData will be available through course website

17

Header

P2

# Created by IrfanView

128 128

255

47

55 58 58 58 58 58 56 56 55 54 53 52

Magic Numbers for format identificationP2 - ASCII PGM P3 – ASCII PPMP5 – Binary PGMP6 – Binary PPM

18

Tasks

Implement both in MATLAB and C/C++MATLAB Introduction

http://math.ucsd.edu/~driver/21d-s99/matlab-primer.html

Tasks

Write functionsReadBinPGM, WriteBinPGM, ReadBinPPM, WriteBinPPM

Input: FileNameOutput: Matrix array containing image data

19

Tasks

1. Read in .pgm image file. Create negative of the image. Save it as another .pgm file

2. Read in a .ppm file. Switch color planes R↔G. Save it as another .ppmfile

3. Read in a .ppm file. Save its R, G, B components as separate .pgm files

Due Date

Tuesday 10th SeptemberCode printout of functionsSample output

MATLAB help?