Computer Vision

Computer Vision
• What is computer vision?
− “Making computers see and understand”

Nice
sunset!

2
 What is Computer Vision?
• Given an image or more, extract properties of the
3D world

• Traffic scene
• Number of vehicles
• Type of vehicles
• Location of closest obstacle
• Assessment of congestion
• Location of the scene
captured
• …
Vision and Computer Vision

4
 Computer Vision

• The computer vision problem can be stated as

follows:
Given a two-dimensional image, infer the objects
that produced it, including their shapes, locations,
colors and sizes.

5
 Vision Transforms From This…
01 00 05 00 03 00 02 00 00 03 01 01 01 01 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 02 00
01
03 30 3A 38 39 2D 1D 15 10 0E 0C 0A 0A 0A 09 06 08 07 06 06 05 05 07 07 04 05 04 04 06 02 01 02 02 02 02 07 01 02 02
03
03 22 1B 16 14 0A 08 0B 0A 0D 0B 0B 0C 06 07 05 05 06 06 06 03 07 04 06 05 09 05 04 05 01 04 04 02 03 03 04 02 04 03
02
00 0F 0B 04 10 07 09 07 08 09 09 08 05 08 08 05 09 03 08 05 02 08 08 06 06 04 02 05 03 02 05 05 00 02 02 04 04 00 00
03
00 07 09 0E 0C 07 08 0A 0A 0B 0F 0A 0C 07 06 0B 07 0B 05 0B 08 09 07 03 08 04 04 02 00 04 02 04 00 04 03 08 00 06 09
04
00 0E 0C 09 09 08 08 07 08 09 09 0A 05 08 07 07 07 09 08 0A 08 09 06 0A 03 09 07 06 06 03 05 03 01 06 02 03 07 01 04
04
02 0C 0B 0A 05 08 09 0A 0C 0A 0A 08 0A 0A 06 08 06 06 04 06 02 06 07 04 04 04 06 09 05 05 08 06 04 05 04 06 01 0A 03
02
02 0B 14 0F 0F 0D 0A 0E 0A 0C 0C 0E 0A 0C 0B 09 0A 09 0A 0A 09 0B 0B 05 0C 0C 0A 04 07 06 03 05 07 04 05 03 02 01 06
03
02 10 12 0B 10 0A 0D 0D 0B 0D 0C 0B 0B 0C 0D 0B 0B 0A 0A 0A 0B 0C 17 15 1C 15 0D 08 09 08 05 05 05 04 02 05 04 04 00
04
01 15 0E 10 12 0C 0D 0C 0C 0A 0B 0B 09 0C 0F 09 09 0D 07 0B 08 15 60 5D 61 59 33 0D 0A 07 08 08 05 03 06 07 01 03 05
02
02 12 10 0F 0E 10 10 0B 0C 0F 0F 0E 0C 10 0D 15 10 09 12 11 12 50 68 66 89 71 5E 3F 08 09 0A 09 0A 03 03 02 05 05 04
02
01 11 12 0C 11 13 10 10 0B 10 0F 0C 11 11 13 0D 0F 0D 0D 0B 25 7A 7F 79 6D 80 6E 54 0C 0D 09 0A 06 04 02 05 00 05 04
03
01 10 0F 0D 12 0E 10 0E 0F 13 13 11 13 17 11 0F 14 11 11 14 39 84 88 7E 8C 73 7A 5C 1E 05 0A 0F 0E 0C 05 02 04 03 06
05
02 0F 15 0D 18 11 0D 11 14 10 12 12 14 19 13 17 13 16 16 20 73 68 87 89 93 8B 83 69 43 07 0A 12 0A 0B 06 06 03 04 05
03
02 13 14 14 16 11 13 13 17 12 17 17 28 1E 1A 17 19 14 12 4F 7D 74 85 91 93 8C 7F 6F 5F 0B 09 12 0D 0C 02 04 07 04 05
04
00 0F 16 0F 13 12 10 1D 12 21 15 1E 21 1F 1C 1D 2D 1A 2D 7C 7A 95 6B 30 48 62 87 71 5C 0A 08 11 0C 09 04 04 02 06 04
03
00 10 1C 10 11 1A 0D 1A 1A 25 28 33 30 26 2B 3E 29 35 6C 83 5E 7B 94 8A 5A 3D 42 76 5C 13 08 13 0F 0C 04 04 01 05 05
03
01 12 17 1A 19 18 15 20 29 20 3F 1F 37 29 39 49 24 33 8F 93 B4 AE 79 42 39 73 7D 89 46 12 06 12 12 0F 08 03 03 03 04
03
01 13 20 0F 14 26 1B 18 20 2F 3D 3E 42 3B 45 2E 48 70 96 9F 96 6B 24 0F 22 4B C3 A4 3F 4F 0C 18 16 0F 05 05 08 05 05
04
00 19 1C 13 13 21 1D 12 18 47 3D 47 45 3A 27 3B 33 A8 A6 91 81 4B A1 75 4B AC A1 B5 79 0C 0B 13 0F 0B 02 03 06 07 07
04
00 1B 1D 1C 1C 1C 1B 1B 1E 55 49 49 36 28 2A 24 9F AD AC AA B1 9C 8D 5F 3E 98 B7 B7 A3 31 11 14 0A 0D 04 08 07 07 07
06
02 21 18 15 16 1D 15 18 1E 36 5B 29 2C 19 29 4F AF BC AF AB 9E A1 97 82 70 9F AE AD A5 92 16 10 07 0E 0A 0C 08 05 0B
05
01 17 1B 1A 1A 2B 1B 2A 32 34 46 2C 1B 26 4C 40 BA BB B5 AE 95 94 84 7A 8A 9A B9 BB AD 9C 8A 15 09 09 05 0B 0D 0F 0B
07 6
00 1A 18 1C 1E 27 21 1D 3F 4E 32 25 1B 1B 93 46 AF AB B1 AC A4 93 89 91 86 90 AA 9F 91 97 AD 7F 0C 0B 0E 0B 0C 0C 09
 …To This
• A harbor with many
dozens of boats; water is
calm and glassy; masts
are all vertical; mountains
in background, blue sky
with a touch of clouds…

7
 …Or To This
• J548043

8
 …Or To This
• Hallway straight ahead

9
 …Or To This
• Angry
Surprised
Happy
Upset

10
 Computer Vision vs. Graphics
• 3D2D implies information loss

graphics

vision

• sensitivity to errors
• need for models
 Why is Computer Vision Difficult?
• It is a many-to-one mapping
− A variety of surfaces with different material and
geometrical properties, possibly under different lighting
conditions, could lead to identical images
− Inverse mapping is under-constrained – non-unique
solution (a lot of information is lost in the transformation
from the 3D world to the 2D image)
• It is computationally intensive
• We do not understand the recognition problem

12
Why is Vision Difficult?

Consider the input...

Not this

13
 But this…
01 00 05 00 03 00 02 00 00 03 01 01 01 01 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 02 00
01
03 30 3A 38 39 2D 1D 15 10 0E 0C 0A 0A 0A 09 06 08 07 06 06 05 05 07 07 04 05 04 04 06 02 01 02 02 02 02 07 01 02 02
03
03 22 1B 16 14 0A 08 0B 0A 0D 0B 0B 0C 06 07 05 05 06 06 06 03 07 04 06 05 09 05 04 05 01 04 04 02 03 03 04 02 04 03
02
00 0F 0B 04 10 07 09 07 08 09 09 08 05 08 08 05 09 03 08 05 02 08 08 06 06 04 02 05 03 02 05 05 00 02 02 04 04 00 00
03
00 07 09 0E 0C 07 08 0A 0A 0B 0F 0A 0C 07 06 0B 07 0B 05 0B 08 09 07 03 08 04 04 02 00 04 02 04 00 04 03 08 00 06 09
04
00 0E 0C 09 09 08 08 07 08 09 09 0A 05 08 07 07 07 09 08 0A 08 09 06 0A 03 09 07 06 06 03 05 03 01 06 02 03 07 01 04
04
02 0C 0B 0A 05 08 09 0A 0C 0A 0A 08 0A 0A 06 08 06 06 04 06 02 06 07 04 04 04 06 09 05 05 08 06 04 05 04 06 01 0A 03
02
02 0B 14 0F 0F 0D 0A 0E 0A 0C 0C 0E 0A 0C 0B 09 0A 09 0A 0A 09 0B 0B 05 0C 0C 0A 04 07 06 03 05 07 04 05 03 02 01 06
03
02 10 12 0B 10 0A 0D 0D 0B 0D 0C 0B 0B 0C 0D 0B 0B 0A 0A 0A 0B 0C 17 15 1C 15 0D 08 09 08 05 05 05 04 02 05 04 04 00
04
01 15 0E 10 12 0C 0D 0C 0C 0A 0B 0B 09 0C 0F 09 09 0D 07 0B 08 15 60 5D 61 59 33 0D 0A 07 08 08 05 03 06 07 01 03 05
02
02 12 10 0F 0E 10 10 0B 0C 0F 0F 0E 0C 10 0D 15 10 09 12 11 12 50 68 66 89 71 5E 3F 08 09 0A 09 0A 03 03 02 05 05 04
02
01 11 12 0C 11 13 10 10 0B 10 0F 0C 11 11 13 0D 0F 0D 0D 0B 25 7A 7F 79 6D 80 6E 54 0C 0D 09 0A 06 04 02 05 00 05 04
03
01 10 0F 0D 12 0E 10 0E 0F 13 13 11 13 17 11 0F 14 11 11 14 39 84 88 7E 8C 73 7A 5C 1E 05 0A 0F 0E 0C 05 02 04 03 06
05
02 0F 15 0D 18 11 0D 11 14 10 12 12 14 19 13 17 13 16 16 20 73 68 87 89 93 8B 83 69 43 07 0A 12 0A 0B 06 06 03 04 05
03
02 13 14 14 16 11 13 13 17 12 17 17 28 1E 1A 17 19 14 12 4F 7D 74 85 91 93 8C 7F 6F 5F 0B 09 12 0D 0C 02 04 07 04 05
04
00 0F 16 0F 13 12 10 1D 12 21 15 1E 21 1F 1C 1D 2D 1A 2D 7C 7A 95 6B 30 48 62 87 71 5C 0A 08 11 0C 09 04 04 02 06 04
03
00 10 1C 10 11 1A 0D 1A 1A 25 28 33 30 26 2B 3E 29 35 6C 83 5E 7B 94 8A 5A 3D 42 76 5C 13 08 13 0F 0C 04 04 01 05 05
03
01 12 17 1A 19 18 15 20 29 20 3F 1F 37 29 39 49 24 33 8F 93 B4 AE 79 42 39 73 7D 89 46 12 06 12 12 0F 08 03 03 03 04
03
01 13 20 0F 14 26 1B 18 20 2F 3D 3E 42 3B 45 2E 48 70 96 9F 96 6B 24 0F 22 4B C3 A4 3F 4F 0C 18 16 0F 05 05 08 05 05
04
00 19 1C 13 13 21 1D 12 18 47 3D 47 45 3A 27 3B 33 A8 A6 91 81 4B A1 75 4B AC A1 B5 79 0C 0B 13 0F 0B 02 03 06 07 07
04
00 1B 1D 1C 1C 1C 1B 1B 1E 55 49 49 36 28 2A 24 9F AD AC AA B1 9C 8D 5F 3E 98 B7 B7 A3 31 11 14 0A 0D 04 08 07 07 07
06
02 21 18 15 16 1D 15 18 1E 36 5B 29 2C 19 29 4F AF BC AF AB 9E A1 97 82 70 9F AE AD A5 92 16 10 07 0E 0A 0C 08 05 0B
05
01 17 1B 1A 1A 2B 1B 2A 32 34 46 2C 1B 26 4C 40 BA BB B5 AE 95 94 84 7A 8A 9A B9 BB AD 9C 8A 15 09 09 05 0B 0D 0F 0B
07 14
00 1A 18 1C 1E 27 21 1D 3F 4E 32 25 1B 1B 93 46 AF AB B1 AC A4 93 89 91 86 90 AA 9F 91 97 AD 7F 0C 0B 0E 0B 0C 0C 09
 Some Possible Outputs

depth object pose action

or (facing away, understanding
segmentation facing forward)

15
 What Is This?
• Texture cues

16
 What Is This?
• Shape cues

17
 What Is This?
• Grouping cues

18
Illusions

19
Illusions

20
Illusions

21
Illusions

22
Illusions

23
 The Three Processing Levels
• Low-level processing
− Standard procedures are applied to improve image quality
− No “intelligent” capabilities

24
 The Three Processing Levels
• Intermediate-level processing
− Extract and characterize components in the image
− Some intelligent capabilities are required

25
 The Three Processing Levels
• High-level processing
− Recognition and interpretation
− Procedures require high intelligent capabilities

26
 Computer Vision
• Some applications

− Robotics
− Navigation, object manipulation, interaction with humans…
− Inspection, measurement
− Medical imaging
− Graphics and animation, special effects
− Multimedia database indexing and retrieval
− Human-computer interaction
− Surveillance and security

27
 Current State of the Art
Earth viewers (3D modeling)

Image from Microsoft’s Virtual Earth

(see also: Google Earth)
Visual Inspection

29
Character Recognition

30
Document Handling
Signature Verification
Biometrics

33
 Target Recognition
• Department of Defense (Army, Airforce, Navy)

34
Interpretation of Aerial Photography

35
 Autonomous Vehicles
• Land, Underwater, Space

36
Traffic Monitoring

37
Inserting Artificial Objects into a Scene

38
Face Detection

39
Face Recognition

40
Human Activity Recognition

41
Login without a password…

Face recognition systems now

Fingerprint scanners on
beginning to appear more widely
many new laptops, http://www.sensiblevision.com/
other devices
 Medical imaging

Image guided surgery

3D imaging
Grimson et al., MIT
MRI, CT