Vision

1
Vision

• To know what is where, by looking. (Marr)

2
Why is Vision Interesting?

• Psychology
• 33 of cerebral cortex for vision
• Vision is how we experience the world
• Engineering
• Want machines to interact with world
• Digital images prevalent

3
Vision is Inferential Light
http//web.mit.edu/persci/people/adelson/checkersh
adow_illusion.html
4
Vision is Inferential Light
http//web.mit.edu/persci/people/adelson/checkersh
adow_illusion.html
5
Vision is Inferential
6
Vision is Inferential Prior Knowledge
7
Computer Vision

• Inference ? Computation
• Building machines that see
• Modeling biological perception

8
A Quick Tour of Computer Vision
9
Boundary Detection Artificial Life
Corpus Callosum deformable organism(G. Hamarneh,
T. McInerney, M. Shenton, D. Terzopoulos)
10
Image Segmentation
11
Tracking
http//www.robots.ox.ac.uk/vdg/dynamics.html
12
Tracking
http//www.robots.ox.ac.uk/vdg/dynamics.html
13
Tracking
14
Tracking
15
Tracking
16
Tracking
17
Tracking
18
Stereo
19
Stereo
Left Image
Right Image
20
Stereo
http//www.magiceye.com/
21
Motion
http//www.ai.mit.edu/courses/6.801/lect/lect01_da
rrell.pdf
22
Motion Application
(www.realviz.com)
23
Pose Determination
Visually-guided surgery
24
Recognition Shading
Lighting affects appearance
25
Recognition Shading
26
Vision depends on

• Geometry
• Physics
• The nature of objects in the world(This is the
  hardest part.)

27
Approaches to Vision
28
Modeling Algorithms

• Build model of world
• Find provably good algorithms
• Experiment on real world
• Update model
• Problem Too often, models simplistic or
  intractable.

29
Bayesian Inference

• Bayes Law P(A B) P(B A) P(A) / P(B)
• P(world image) P(image world) P(world) /
  P(image)
• P(image world) is computer graphics
• Geometry of projection
• Physics of light reflection, refraction, etc.
• P(world) involves modeling objects in world
• Leads to statistical/learning approaches
• Problem Too often, probabilities unknown and
  thus invented.

30
Engineering

• Focus on concrete tasks with clear requirements
• Test ideas based on theory and gain experience
  about what works
• Try to build reusable modules
• Problem Solutions that work under specific
  conditions may not generalize.

31
State of Computer Vision Science

• Study of intelligence seems hard
• Several interesting fundamental theories about
  specific problems
• Limited insight into how these theories interact

32
State of Computer Vision Technology

• Interesting applications security, visual
  effects, compression, photography
• Some successful companies. Largest 100-200
  million in revenues. Many in-house applications.
• Future growth in digital imagery exciting

33
Related Fields

• Graphics Vision is inverse graphics
• Visual perception, neuroscience
• Learning, optimization
• Math e.g. geometry, stochastic processes