Tony Pridmore
Room C57
tpp@cs.nott.ac.uk
Computer Vision
An image is a function of
- the reflectance and shape of the viewed object
- illumination conditions
- viewing geometry
Biological vision inverts that function
The goal of computer vision is to
- understand how
- build systems that do something similar and useful
Computer Vision
Vision is underdetermined
- one pixel value is not enough to allow us to solve for
everything we need
Vision is overdetermined
- multiple cues are available; shading, texture, stereo,
motion, boundary shape,…..
Vision is impossible without assumptions/prior knowledge
G52IVG and G53VIS
Companion to G52IVG, which introduced some general
problems and techniques in computer vision and graphics….
- Image Acquisition and Representation
- Image Filtering
- Edge Detection & Hough Transforms
- Colour & Segmentation
- Motion & Tracking
- Projective Geometry
G52IVG and G53VIS
….. and gave some practical experience of implementing
image processing and vision techniques in JAVA
http://www.cs.nott.ac.uk/~bai/IVG/practicals.html
If you haven’t done the IVG practicals, now would be a
good time
G52IVG and G53VIS
G53VIS will revisit and build upon some of the topics
covered in G52IVG
- more advanced/powerful/useful techniques
- some new problems/methods
- some real vision/image analysis applications
Two lectures/week: Monday 4-5, A26, Business School
South , Tuesday1-2 in B18, Amenities Building
One tutorial/week: Friday 12-1, A24, Business School South
www.cs.nott.ac.uk/~tpp/G53VIS/G53VIS.html
Segmentation
Graph-based methods: Normalised Cuts
Segmentation
Edge-based methods:
Watershed Algorithms
Pixel Clustering Methods:
- K Means
- Expectation Maximisation
Evaluating Segmentation
Algorithms
Binocular Stereo
Motion & Tracking
Optic Flow
Motion & Tracking
Tracking: Kalman Filters
Motion & Tracking
Tracking: Particle filters and Snakes
Multi-target Tracking