CS 4495/7495 Computer Vision
Fall 2000 Syllabus

  1. Introduction
      1. Applications and our research
      2. Image Processing vs Computer Vision

I) Images and Pixels

  1. Image formation

a.       Geometry

                                                               i.      Ideal perspective projection

                                                             ii.      Hint of stereo

                                                            iii.      Real lenses and cameras

b.      Image as continuous function

                                                               i.      Example linear operations - convolution

                                                             ii.      General image properties: brightness, contrast frequency

                                                            iii.      Noise vs signal - SNR

                                                           iv.      Example: SNR vs Contrast

Digression: Human vision

c.       Digitization

                                                               i.      Sampling in space

                                                             ii.      Quantization, dynamic range

                                                            iii.      Integer vs continuous values

d.      Vector images

                                                               i.      color

                                                             ii.      z-buffer

c.       Pre-processing

a.       Global Photometric

                                                               i.      Gain normalization

                                                             ii.      Histogram equalization

b.      Geometric

                                                               i.      Re-sampling

                                                             ii.      Sub-sample vs interpolation

                                                            iii.      Interpolation methods

c.       Pyramids

                                                               i.      Gauss reduction -theory

                                                             ii.      Nyquist rates

                                                            iii.      Burt pyramid

                                                           iv.      Quad trees

d.      Deformations/mappings

                                                               i.      Notion of operator

                                                             ii.      Planar models: Affine, perspective

e.       Filtering

                                                               i.      Blurring

                                                             ii.      Other local operators - reduce information

                                                            iii.      Non-linear: median

                                                           iv.      Structure sensitive enhancement (Freeman)

f.        Edge detection

                                                               i.      Canny, Marr &Hildreth,

d.      Binary Vision

a.       Generation

                                                               i.      Thresholds – absolute and dynamic

Digression: Background subtraction

b.      Binary images, morphology

                                                               i.      dilation and erosion

                                                             ii.      Morphological Matching

                                                            iii.      skeletons

c.       Region properties

                                                               i.      Euler numbers,

                                                             ii.      Moments

e.       Finding structure

a.       Hough transform

                                                               i.      Lines

                                                             ii.      Generalized

b.      RANSAC

c.       Other machine vision techniques

d.      Case study: Face recognition

f.        Image motion

a.       Optic flow - brightness constraint

                                                               i.      Kanade etc method

                                                             ii.      Hierarchical algorithms

                                                            iii.      Parametric methods

II Beyond pixels: Seeing the world

g.       3D Vision

a.       Projective geometry

                                                               i.      Calibration

                                                             ii.      Essential and Fundamental matrices

b.      Stereo

                                                               i.      Random-dot

                                                             ii.      SSD

                                                            iii.      Correlation, occlusion, DP

c.       Radiometry and 3D vision

                                                               i.      Surface reflectance

                                                             ii.      Shape from shading

                                                            iii.      Photometric stereo

h.       Camera and structure motion

a.       Egomotion in velocity space

b.      Recovery of egomotion

                                                               i.      Translation only - FOE

                                                             ii.      Full - Heeger and Jepson??

i.         Model-based recognition

a.       3D models

                                                               i.      Goad's algorithm

                                                             ii.      Interpretation trees

b.      Class based models

                                                               i.      Active shape models

j.        Video Analysis

a.       Representing video

                                                               i.      Layers

                                                             ii.      MJPEG, MPEG

b.      Tracking

                                                               i.      Kalman methods

                                                             ii.      CONDENSATION-like methods

                                                            iii.      Structural hacks

c.       Sequence analysis

                                                               i.      HMMs

                                                             ii.      Gesture recognition

d.      Watching people

                                                               i.      Limb tracking with weak models

                                                             ii.      Limb tracking with strong models