BMHA308L Machine Vision

Dr. C. Moganapriya -- Machine Vision

Syllabus Overview

Dr. C. Moganapriya -- Machine Vision

 Digital Assignments
• Woking model
• Maximum three per team
• 0+ Three reviews- each 10 marks
• CA 1: Unit 1, 2; CA2: 3,4,5

Attendance: No late comers will be encouraged for excuses.

Dr. C. Moganapriya -- Machine Vision

 Guest Lecture
Mr. Prabhu, AGM, Manufacturing,
Ford India Pvt. Ltd

Dr. C. Moganapriya -- Machine Vision

Why this subject?

Dr. C. Moganapriya -- Machine Vision

Real Time Application: Coin Counting

Dr. C. Moganapriya -- Machine Vision

Image Processing

Dr. C. Moganapriya -- Machine Vision

Image Processing contd…

Dr. C. Moganapriya -- Machine Vision

Image Processing contd…

Dr. C. Moganapriya -- Machine Vision

Image Processing contd…

Dr. C. Moganapriya -- Machine Vision

Image Processing contd…

Dr. C. Moganapriya -- Machine Vision

Image Processing contd…

Dr. C. Moganapriya -- Machine Vision

Image Processing – Background
Elimination

Dr. C. Moganapriya -- Machine Vision

Image Processing – Edge Detection

Dr. C. Moganapriya -- Machine Vision

Image Processing – Binarization using
Thresholding

Dr. C. Moganapriya -- Machine Vision

Image Processing - Segmentation

Dr. C. Moganapriya -- Machine Vision

Coin Counting

Dr. C. Moganapriya -- Machine Vision

Coin Counting

Dr. C. Moganapriya -- Machine Vision

Coin Counting

Dr. C. Moganapriya -- Machine Vision

Coin Counting

Dr. C. Moganapriya -- Machine Vision

Coin Counting

Dr. C. Moganapriya -- Machine Vision

Coin Counting

Dr. C. Moganapriya -- Machine Vision

Coin Counting

Dr. C. Moganapriya -- Machine Vision

Coin Counting

Dr. C. Moganapriya -- Machine Vision

Coin Counting

Dr. C. Moganapriya -- Machine Vision

Coin Counting
19
rupees

Dr. C. Moganapriya -- Machine Vision

Coin Counting
19
rupees

Dr. C. Moganapriya -- Machine Vision

Edibility Analysis of Spinach Leaves

Dr. C. Moganapriya -- Machine Vision

IMAGE PROCESSING

• Enhancing an image

• Extracting information or features from an image

• To obtain categories of information from specific features.

Dr. C. Moganapriya -- Machine Vision

IMPORTANT STAGES
IN IMAGE PROCESSING

Dr. C. Moganapriya -- Machine Vision

IMPORTANT STAGES
IN IMAGE PROCESSING

Coin
counting

Dr. C. Moganapriya -- Machine Vision

IMPORTANT STAGES
IN IMAGE PROCESSING

Coin
counting Acquire/
img db

Dr. C. Moganapriya -- Machine Vision

IMPORTANT STAGES
IN IMAGE PROCESSING

Img
enhancem
ent / Noise
removal

Coin
counting Acquire/
img db

Dr. C. Moganapriya -- Machine Vision

IMPORTANT STAGES
IN IMAGE PROCESSING

Img
enhancem
ent / Noise
removal

Coin
counting Acquire/
img db

Dr. C. Moganapriya -- Machine Vision

IMPORTANT STAGES
IN IMAGE PROCESSING

Img
enhancem
ent / Noise
removal

Coin
counting Acquire/
img db

Dr. C. Moganapriya -- Machine Vision

IMPORTANT STAGES
IN IMAGE PROCESSING

5 rupee
coin
Img
enhancem
ent / Noise
removal

Coin
counting Acquire/
img db

Dr. C. Moganapriya -- Machine Vision

IMPORTANT STAGES
IN IMAGE PROCESSING

5 rupee
coin
Img
enhancem
ent / Noise
removal

Coin
counting Acquire/
img db

Dr. C. Moganapriya -- Machine Vision

IMPORTANT STAGES
IN IMAGE PROCESSING

5 rupee
coin
Img
enhancem
ent / Noise
removal

Coin
counting Acquire/
img db

Dr. C. Moganapriya -- Machine Vision

IMPORTANT STAGES
IN IMAGE PROCESSING

5 rupee
coin
Img
enhancem
ent / Noise
removal

Coin
counting Acquire/ Non-coin
img db 2 rupee 5 rupee

defective
Dr. C. Moganapriya -- Machine Vision
IMPORTANT STAGES
IN IMAGE PROCESSING

• Image Acquisition
• Preprocessing
• Segmentation
• Representation and Description
• Recognition and Interpretation
• Knowledge base

Dr. C. Moganapriya -- Machine Vision

Eye and Human Vision

Dr. C. Moganapriya -- Machine Vision

Content
1. Is human vision a good model for computer vision?
2. How does human vision work (and how does it fail)?
3. Software languages & associated literature

Dr. C. Moganapriya -- Machine Vision

Modelling the eye in three parts

Physical Experimental Psychological

model model model
(the eye) (the signals) (the brain)

Each is not fully understood, especially the brain ……

Dr. C. Moganapriya -- Machine Vision

Human eye

Evolved for survival

Function of the eye is to form an image
on the retina (on fovea)
The lens is shaped, rather than moved
Image is transmitted via optic nerve

Dr. C. Moganapriya -- Machine Vision

Optics
Your brain must invert the image

Dr. C. Moganapriya -- Machine Vision

[Link]
Sensors
There must be a lot!
Cones (107) and rods (108)
Cones – colour; rods – greylevel
photopic scotopic
Cones come in three types
1. S – short wavelength (blue)
2. M – medium wavelength (green)
3. L – long wavelength (red)
Insufficient bandwidth of optic nerve
implies coding

Dr. C. Moganapriya -- Machine Vision

Rod and cone densities

No sensors on blind spot

Most cones on fovea
Rods elsewhere

Dr. C. Moganapriya -- Machine[Link]

Vision
 [Link]
Dr. C. Moganapriya -- Machine Vision content/uploads/2011/03/[Link]
Spectral responses

Blue response (S sensors) is poor

Dr. C. Moganapriya -- Machine Vision

Spectral responses

Blue response (S sensors) is poor

Green response (M sensors) dominates

Dr. C. Moganapriya -- Machine Vision

Spectral responses

Blue response (S sensors) is poor

Green response (M sensors) dominates
Red response (L sensors) close to heat

Dr. C. Moganapriya -- Machine Vision

Spectral responses

Blue response (S sensors) is poor

Green response (M sensors) dominates
Red response (L sensors) closer to heat
Overall response from S, M and L
Human vision must
compensate for the poor
blue sensors
(Just like a camera)

Dr. C. Moganapriya -- Machine Vision

[Link]
k/NBSummer15/NBSummerLect4.
html Dr. C. Moganapriya -- Machine Vision
Cortices

Dr. C. Moganapriya -- Machine Vision

[Link]
Neural processing

Sensor
information must
be combined

Note Weber’s law

Dr. C. Moganapriya -- Machine Vision

Where are we?

links

patterns
Dr. C. Moganapriya -- Machine Vision
HUMAN VISUAL PERCEPTION

57
HUMAN VISUAL PERCEPTION

58
BRIGHTNESS ADAPTATION & DISCRIMINATION

• The eye also discriminates between changes in

brightness at any specific adaptation level.

I c
→ Weber ratio
I
Where: Ic: the increment of illumination
discriminable 50% of the time and
I : background illumination

59
BRIGHTNESS ADAPTATION & DISCRIMINATION

• Small values of Weber ratio mean good brightness discrimination

(and vice versa).

• At low levels of illumination brightness discrimination is poor (rods)

and it improves significantly as background illumination increases
(cones).

60
BRIGHTNESS ADAPTATION & DISCRIMINATION

• The typical observer can discern one to two dozen different intensity
changes

• i.e. the number of different intensities a person can see at any one point in a
monochrome image

61
Dr. C. Moganapriya -- Machine Vision
BRIGHTNESS ADAPTATION & DISCRIMINATION

• Overall intensity discrimination is broad due to different set of

incremental changes to be detected at each new adaptation level.

• Perceived brightness is not a simple function of intensity

• Scalloped effect, Mach band pattern
• Simultaneous contrast

63
• RADIANCE is the total amount of energy that flows from the light
source
• LUMINANCE gives the measure of the amount of energy an observer
perceives from a light source.
• BRIGHTNESS is a subjective descriptor of light perception that is
practically impossible to measure.

64
PERCEIVED BRIGHTNESS

65
SIMULTANEOUS CONTRAST

66
Mach bands

Mach bands are not in the image:

your vision introduces them

Result of brightness adaption

Dr. C. Moganapriya -- Machine Vision

How human vision uses edges
The human eye needs training and can be deceived

Dr. C. Moganapriya -- Machine Vision

Static illusions
Measurement needs comparison

Dr. C. Moganapriya -- Machine Vision

Dr. C. Moganapriya -- Machine Vision
Dr. C. Moganapriya -- Machine Vision
Benham’s disk
Illusions are a consequence of complex function

Dr. C. Moganapriya -- Machine Vision

ILLUSIONS

73
ELEMENTS OF VISUAL PERCEPTION OF THE IMAGE

CONTRAST

BORDER SHAPE

VISUAL PERCEPTION

BRIGHTNESS
BACKGROUND

TEXTURE

74
 CONTRAST
Objects in an image can be identified only when it is distinguished
from its background

CONTRAST CONTRAST CONTRAST

Local change in the brightness of and object and the background.

Higher the brightness , more is the contrast.
Brightness depends on the brightness of the local surroundings
– conditional contrast

LUMINANCE it describes the amount of light that passes

through or is emitted from a particular area and falls within a
given solid angle.

75
 ACUITY
Ability to detect details in an image.

Human eye – sensitive to slow and fast changes

76
 BEST RESOLUTION

60wt

400mm
500lux

   250mm 
77
 BORDER

Object borders carry a lot of information for humans.

It enables adaptation effects - Ebbinghaus illusion.

78
 PERCEPTUAL GROUPING

To aggregate elements provided by low-level operations –

grouping according to

Properties of elements

parallelism
symmetry
continuity
Closure
not grouped
Proximity
Similarity
direction
79
Not grouped

Proximity

Similariy

80
 D

1. Closure
2. Parallelism
3. Symmetry
4. Continuity

C 81
Digital Image
DIGITAL IMAGES are electronic snapshots taken of a scene or scanned
from documents, such as photographs, manuscripts, printed texts, and
artwork.
•A digital image is a grid of tiny squares called pixels.
•Each pixel stores color or brightness data as numbers.

82
DIGITAL IMAGE

1 pixel

83
IMAGE FORMATION

Digital Image can also be called as conversion of one dimensional

energy/signal to two dimensional representation

f(x,y) = reflectance(x,y) * illumination(x,y)

84
IMAGE MODEL

f(x,y) = reflectance(x,y) * illumination(x,y)

F(x)= f(x,y) where f(x,y) is the brightness of the graylevel

of the image at a spatial coordinate (x,y)
85
DIGITIZED IMAGE
f(x,y) ie
fred(x,y), fgreen(x,y), fblue(x,y)

86
DIGITIZED IMAGE
f(x,y) ie f(2,0) =0
f(x,y) ie f(2,0) =128

87
Dr. C. Moganapriya -- Machine Vision
DIGITIZED BINARY IMAGE
1 1 0 1 1
1 0 1 0 1
1 1 0 1 1
0 0 0 0 0
1 1 0 1 1
1 1 0 1 1
1 0 1 0 1
0 1 1 1 0

89
Bit Depth
.
• No. of bits used to indicate the color of a single pixel
• The higher the bit depth of an image, the more colors it can store.
• 1 bit image, can only show two colors, black and white
• 1 bit can only store one of two values, 0 (white) and 1 (black).

90
Bit Depth

91
Bit Depth

92
Bit Depth

93
Bit Depth

94
Bit Depth

95