Computer Science 153
Computer Vision
Lecture notes and schedule, Fall 2000
On This Page
Lecture Slides
Tentative Lecture Schedule
Lecture Slides (Microsoft PowerPoint)
Lecture 1 slides (Fall 00)
Lecture 2 slides (Fall 00)
Lecture 3 slides (Fall 00)
Lecture 4 slides (Fall 00)
Lecture 5 slides (Fall 00)
Color Constancy Demo #1
Color Constancy Demo #2
Lecture 6 slides (Fall 00)
Only the first step...
Set Demo #1
Lecture 7 slides (Fall 00)
Lecture 8 slides (Fall 00)
Lecture 9 slides (Fall 00)
Lecture 10 slides (Fall 00)
Lecture 11/12 slides (Fall 00)
Lecture 12 slides (Fall 00)
Lecture 13 slides (Fall 00)
Lecture 14 slides (Fall 00)
Lecture 15 slides (Fall 00)
Lecture 16 slides (Fall 00)
Lecture 17 slides (Fall 00)
Lecture 18 slides (Fall 00)
Lecture 19 slides (Fall 00)
Lecture 20 slides (Fall 00)
Lecture 21 slides (Fall 00)
Lecture 22 slides (Fall 00)
Lecture 23 slides (Fall 00)
Lecture 24 slides (Fall 00)
Lecture 25 slides (Fall 00)
Lecture Schedule
This list constitutes the basic topics of CS153; they
will be the focus of the lectures, assignments, and exams
for the course. Please keep in mind that this is tentative... .
- (8/29) - What is Computer Vision?
- The problem
- The solution?
- Explaining the first assignment
- A bit about the course in general
- (8/31) - Matlab and experimenting with images
- An image as a matlab array
- Getting at parts of images: subimages, lists of pixels
- Introduction to the tools in the graphics lab
- Assignment 1 due Monday, September 4 by midnight.
- Solving the vision problem.
- (9/5) - A statistical approach to vision
- Classifying pixels: greyscale and color
- How to describe light/dark or color: colorspaces
- Image formats
- Other ways to decompose images: Fourier analysis
- Properties and problems in images: quantization, sampling, aliasing,
noise, ...
- (9/7) - Thresholds and histograms
- Creating binary images: segmentation
- Finding binary image regions: connected components
- 4 vs. 8 connectedness
- Applying image masks and general morphology
- (9/12) - Pattern Recognition
- K-means analysis
- likelihood classifiers
- Nondeclarative means (n nets, genetic algorithms)
- (9/14) - More pattern recognition
- Content-based image retrieval
- Confusion matrix
- Other features used: texture
- unoriented vs. oriented (edges)
- Assignment 2
- Using histograms for recognition
- Segmentation & Matching
- (9/19) - Texture
- Co-occurence matrix
- Image entropy
- Edges as a particular kind of local texture
- (9/21) - Color
- Color representations
- Comparing color content
- Histogram intersection
- Weighted histogram distance
- Earth-mover's distance
- Color constancy
- (9/26) - Edge Detection and analysis
- What are edges?
- How to detect edges
- Scale
- Convolution
- (9/28) - Collections of edges
- Thinning
- Grouping
- Hough Transform
- Shape representation
- General segmentation algorithms
- (10/3) - More feature detection
- Hough transform - circles
- Templates and correlation vs. convolution
- General purpose filters
- What are good features?
- Assignment 3
- Segmentation using color, texture, edges: jigsaw puzzle
- Histogram and feature matching for object recognition
- Clustering parameter vectors to "learn" objects
- (10/5) - Recognition by alignment
- Geometry of image formation
- 6d pose space
- Narrowing possible space via feature matching
- (10/10) - Recognition by alignment
- Orthographic, Affine, Projective models
- Issues in alignment
- (10/12) - Laboratory work on Assignments
- Matlab's compiler: creating MEX files
- File I/O neede for reading image DB
- Final Project Proposal
- Idea, some background, and a planned approach
- (10/17) - Fall Break
- (10/19) - Shape from X
- Shading
- Photometric stereo
- multiple images: motion, stereo
- (10/24) - More shape from X
- Sensitivity of the reconstruction algorithms
- Regularization: assuming the world is a "nice" place
as a means of adding necessary constraints
- (10/26) - Physics-based vision -> Vision-based vision
- Review of alignment and shape from X
- Introduction to appearance-based modeling and
recognition
- Assignment 4
- Recognition and pose determination of objects
- (10/31) - Appearance-based approaches
- SSD/template matching
- An image as one big vector
- Eigenspaces
- (11/2) - Appearance-based approaches
- SVD
- Eigenfaces vs. Fischerfaces
- other eigenspaces (histograms)
- (11/7) - Learning a visual representation
- Neural nets and Hidden Markov Models: training & using
- (11/9) - Problems with multiple images
- Correspondence
- Introduction to motion and stereo
- Assignment 5
- Face recognition through appearance-based modeling
- (11/14) - Stereo vision
- Geometry
- Depth estimation
- Finding correspondences
- Epipolar constraint
- RANSAC procedure
- (11/16) - Calibrating cameras
- Projective geometry
- Single-camera calibration
- Stereo calibration
- Scene-based multiple-view reconstruction
- (11/21) - Multiple views through motion
- Motion detection
- Image differencing
- Optical flow
- (11/23) - Thanksgiving Recess
- (11/28) - Token Tracking
- Good features to track
- Tracking vs. feature recognition
- Extra Credit Assignment
- (11/30) - Tracking Demonstrations
- XVision
- Primitives: lines, intensity patches (SSD tracking)
- Composite trackers: corners, contours
- Data Association filters and condensation
- (12/5 & 12/7) - Project meetings
- Preliminary student reports and case studies of
comparable vision systems in the field.
- Out-of-class project meetings with students
- Final project presentation - Dec. 10