Computer Science Lecture

1

The U

niv

ers

ity

of M

ancheste

r

COMP27112Computer Graphics and Image Processing

Lecture B1

Introduction to Image Processing

The U

niv

ers

ity

of M

ancheste

r

Contact Details

• Room 2.107

• [email protected]

• 63376

2

The U

niv

ers

ity

of M

ancheste

r

Outline of Lectures

• Image Representation

• Point Processing

• Region Processing (2 Weeks)

• Image Processing

HistoryWhat can we doResolutionColour

The U

niv

ers

ity

of M

ancheste

r

Outline of Lectures





Grey value transformsHistogramThresholdingZooming

Co-Ordinate transformsAffineNon-Affine

3

The U

niv

ers

ity

of M

ancheste

r

Outline of Lectures





ConvolutionSmoothingEdge detectionTemplate MatchingRank Order Filters

The U

niv

ers

ity

of M

ancheste

r

Outline of Lectures




• Image Processing Region DescriptionFinding BlobsMeasurements

4

The U

niv

ers

ity

of M

ancheste

r

Handout

• Backs up these slides

– More material

– More detailed

The U

niv

ers

ity

of M

ancheste

r

Outline of Lab Schedule

• Coursework Assignments

– A mark for submission

– Unsupervised

• Laboratory classes

– 5% per submission

– Supervised

• Details in handouts on the website

5

The U

niv

ers

ity

of M

ancheste

r

Lecture B1

• Definition

• Image Processing History and Examples

• Representation

• Colour

The U

niv

ers

ity

of M

ancheste

r

What is Image Processing?

6

The U

niv

ers

ity

of M

ancheste

r

Historical Digital Images

• 1921

• First digital images, for telegraphic transmission

• Five grey levels

• 3 hours to transmit

The U

niv

ers

ity

of M

ancheste

r

Early Digital Image Processing

• 1960s

• Tied to development of hardware and software

• Space exploration

– to correct for lens distortion

7

The U

niv

ers

ity

of M

ancheste

r

Data Sources

• Anything that can generate a spatially coherent measurement of some property can be imaged

• Can use many energy sources

• Electromagnetic

� γ rays, X rays, visible, microwave, radio

• Others

– sound, magnetic fields

The U

niv

ers

ity

of M

ancheste

r

γ ray images

8

The U

niv

ers

ity

of M

ancheste

r

X ray images

The U

niv

ers

ity

of M

ancheste

r

UV

9

The U

niv

ers

ity

of M

ancheste

r

Visible, IR

The U

niv

ers

ity

of M

ancheste

r

Microwave and Radar

10

The U

niv

ers

ity

of M

ancheste

r

LIDAR

The U

niv

ers

ity

of M

ancheste

r

Applications

• In many areas

– for mapping

– exploration

– recording

– measurement

11

The U

niv

ers

ity

of M

ancheste

r

Medicine

The U

niv

ers

ity

of M

ancheste

r

12

The U

niv

ers

ity

of M

ancheste

r

Oil Exploration

The U

niv

ers

ity

of M

ancheste

r

Astronomy

13

The U

niv

ers

ity

of M

ancheste

rT

he U

niv

ers

ity

of M

ancheste

r

14

The U

niv

ers

ity

of M

ancheste

rT

he U

niv

ers

ity

of M

ancheste

r

15

The U

niv

ers

ity

of M

ancheste

r

Weather

Katrina

The U

niv

ers

ity

of M

ancheste

r

Industry

16

The U

niv

ers

ity

of M

ancheste

r

Security

The U

niv

ers

ity

of M

ancheste

r

Satellite Images

17

The U

niv

ers

ity

of M

ancheste

r

Satellite Images

The U

niv

ers

ity

of M

ancheste

r

Face Recognition – Biometric Passports

You recognise faces

Why not automate?

Passport stores measurements of your face

Scanner captures an image and replicates the measurements

If they match – you’re in

18

The U

niv

ers

ity

of M

ancheste

r

Image Capture

• Many sources

– So we’ll ignore them

– Ultimately they deliver a digital image

• A rectangular array of integer values

• Resolution

– Array dimensions

• Related to the scene

– The “integer values”

The U

niv

ers

ity

of M

ancheste

r

Image Resolution

• How many pixels?

– Spatial resolution

• How many shades of grey/colours?

– Amplitude resolution

• How many frames per second?

– Temporal resolution

• Nyquist’s theorem

19

The U

niv

ers

ity

of M

ancheste

r

Spatial Resolution

n, n/2, n/4, n/8, n/16 and n/32 pixels on a side.

The U

niv

ers

ity

of M

ancheste

r

Field of View

NumberOf Pixels

Resolution = x⁰ per pixel

20

The U

niv

ers

ity

of M

ancheste

r

Nyquist’s Theorem

• A periodic signal can be reconstructed if the sampling interval is half the period

• An object can be detected if two samples span its smallest dimension

The U

niv

ers

ity

of M

ancheste

r

Sine wave

21

The U

niv

ers

ity

of M

ancheste

r

Amplitude Resolution

• Humans can see:

– About 40 shades of brightness

– About 7.5 million shades of colour

• Cameras can see:

– Depends on signal to noise ratio

– 40 dB equates to about 20 shades

• Images captured:

– 256 shades

The U

niv

ers

ity

of M

ancheste

r

Shades of Grey

256, 16, 4 and 2 shades.

22

The U

niv

ers

ity

of M

ancheste

r

Colour Representation

• Newton

– White light composed of seven colours

• red, orange, yellow, green, blue, indigo, violet

• Young etc.

– Three primaries could approximate many colours

• red, green, blue

– CIE

• Define three primary colours: x, y, z (reddish, greenish, blueish)

• These match HVS

The U

niv

ers

ity

of M

ancheste

r

CIE Primaries

0

0.5

1

1.5

2

2.5

350 450 550 650 750

sen

sit

ivit

y

wavelength

CIE colour matching

Series1

Series2

Series3

23

The U

niv

ers

ity

of M

ancheste

r

CIE Colour Diagram

The U

niv

ers

ity

of M

ancheste

r

Other Colour Models

• YCrCb

– Intensity and Colour difference

• Used in broadcasting

• Perceptual Spaces

– HSV | IHS | HSB

– Lab

• All separate intensity/brightness and chromaticity

24

The U

niv

ers

ity

of M

ancheste

r

YCrCb

• Part of the standard for digital video

• Less detail in Cb, Cr so fewer pixels stored

• One Cb and one Cr pixel per four Y pixels

1280813.04187.05.0

1285.03313.01687.0

114.0587.0299.0

+−−=

++−−=

++=

BGRCr

BGRCb

BGRY

The U

niv

ers

ity

of M

ancheste

r

Why Perceptual Spaces?

• Equal distances on CIE diagram DO NOT correspond to equal changes in perceived colour

• This is important for measurement

25

The U

niv

ers

ity

of M

ancheste

r

IHS

• Intensity

– Aka Brightness

– Weighted average of RGB

• Hue

– The basic colour

– An angle from 0 to 359

• Saturation

– Depth of colour

The U

niv

ers

ity

of M

ancheste

r

Lab

• Also known as L*a*b* and CIELAB

– Used in Photoshop

• Can represent all perceivable colours

– Because it’s derived from CIE XYZ co-ordinates

• Device independent

• L* represents lightness

– 0 = black, 100 = white

• a* represents green � magenta

• b* represents blue � yellow

26

The U

niv

ers

ity

of M

ancheste

r

Summary

• Chapters 1 and 2 of handout

• Sample applications

• Resolution

• Colour models

The U

niv

ers

ity

of M

ancheste

r

640k ought to be enough for anybody

Bill Gates, 1981

Computer Science Lecture

Documents