What is an Image?

image
a visual representation of some measurable property of a 4/10/14  
source, object, or phenomenon

Lecture  4  

Digital  Image  Processing  

What is an Image?
Most commonly defined as a two dimensional
What  
function, f(x,y)=I, where x andis  yaare
n   I mage?  
spatial
and I is the grey level or intensity of the image at that
coordinates
point. x

1  
3
 4/10/14  

9/8/2009

What  is  an  Image?  

What is an Image?
Image Surfaces.
Continuous or discrete domain

9/8/

Image  Acquisi=on  
Image Acquisition

2  
 4/10/14  

Sampling  &  Quan=za=on  

Sampling and Quantization

9/8/2009

Sampling  
Sampling &  quan=za=on  
and Quantization

3  

Sampling and Quantization

 4/10/14  

Bit  Depth  
(I) Counting Gray Levels: Factors of 2 (bits)

Number of
One pixel that can display
Levels 2n bits
256 different gray levels
2 21 1
is an 8 bit pixel.
4 22 2
8 23 3
8 bits = 1 byte
16 24 4
32 25 5
One pixel that can display
64 26 6
256 different gray levels
128 27 7
is a 1 byte pixel.
256 28 8

Imaging Science Fundamentals Chester F. Carlson Center for Imaging Science

Thenumber
  The
  numberof
ofpossible
possiblegray
graylevelsBit  Depth  
levelsisiscontrolled
controlled
bythe
by thenumber
numberof
ofbits/pixel,
bits/pixel,or
orthe
the‘bit
‘bitdepth’
depth’of
of
theimage
the image
bits/pixel bits/pixel
11
22
33
44
55
66
77
88
Imaging
ImagingScience
ScienceFundamentals
Fundamentals Chester
ChesterF.F.Carlson
CarlsonCenter
Centerfor
forImaging
ImagingScience
Science

4  
 4/10/14  

9/8/20099/8/2009

DIP  
Example
Example ––MRI MRI ofof Knee Knee
Example:  MRI  of  the  knee     9/8/2009
Problem
ProblemStatement Statement
 Medical images are read by radiologists
 Medical
 Opinionsimages
may varyare
fromread by radiologists
radiologist to radiologist
Opinions
 Changes may
over timevary from radiologist
are difficult to track to radiologist
Changes
 Early over
changes aretime are difficult
difficult to detect to track
Example – MRI of Knee
 Computer-generated assessment
 Early changes are difficult of medical images
to detect
Problem
 Imaging assays are objective,
 Computer-generated Statement
precise
assessment and fully reproducible
of medical
 Imaging assays involve numerical measurement of structure,
images
 Imaging assays are objective,
 Medical images precise
are readand fully reproducible
by radiologists
composition, skeletal alignment or other clinically significant
Imaging assays involve
 factors Opinionsnumerical measurement
may vary from of structure,
radiologist to radiologist
composition,
 Imaging skeletal
assays complimentalignment
 Changes over ordifficult
time are
radiological other
reads, clinically
to track
speed significant
up the reading
factors
and  Early
make them more changes are difficult to detect
accurate.
 Imaging assays compliment radiological
 Computer-generated reads,of
assessment speed up the
medical reading
images
and make them more accurate.
 Imaging assays are objective, precise and fully reproducible
 Imaging assays involve numerical measurement of structure,
composition, skeletal alignment or other clinically significant
factors
 Imaging assays compliment radiological reads, speed up the reading
and make them more accurate.
Image
Acquisition
Example – MRI of Knee
Approach Image
Image
Acquisition
Example – MRI
 Specific aims ofMRI   ooff  Knee
this Example the  knee  
Enhancement

Approach
 Segment knee bone and cartilage
Example – MRI of Knee
Registration Image
Image
of Acquisition
Image toEnhancement
Atlas
 Provide information on describing
 Specific
deviationaims of displacement
this Example
Approach
of bone from Image
Enhancement
normal.
Segment knee bone and cartilage Segment
Registration of
 Specific aims of this Example
 Provide information describing contact
Image
Image to Atlas
 Provide information on describing
 Segment
points between femur and knee bone and cartilage
tibia Registration of
deviation ofbone
Provide
displacement
information
from
on describing
Analyze
Image to Atlas
cartilage
normal. deviation of bone displacement fromResults Segment
ImageSegment
normal. describing contact
 Provide information
Image
 Provide
points between femur information
and tibiadescribing contact
Present
cartilage points between femur and tibia FindingsAnalyze
cartilage ResultsAnalyze
Results

PresentPresent
Findings
Findings
7

7
7

5  
 4/10/14  
9/8/2009

9/8/2009

Example –Image  
MRI ofAcquisi=on  
Knee
Acquisition
Example – MRI of Knee
Acquisition

Example – MRI of Knee

ExampleImage  
Acquisition – MRIAcquisi=on  
of Knee
Acquisition

88

6  
 9/8/2009

4/10/14  

9/8/2009

Example – MRI of Knee

Image  Enhancement  
Image Enhancement
Example – MRI of Knee
Image Enhancement

Image  
Example
Example S–
egmenta=on  
–MRI
MRIof
of Knee
Knee&  Classifica=on    
Registration
Registration

7  
9
 9/8/2009
4/10/14  
9/8/2009

9/8/2009

Example – Image  
MRI of Knee
Example – MRIRofegistra=on  
Knee
Segmentation &&
Segmentation Classification
Classification
Example – MRI of Knee 9/8/2009

Segmentation & Classification

Example – MRI of Knee

Segmentation & Classification

Example – MRI of Knee

Example Data  of
Presentation
– MRI Visualiza=on  
Knee
Example – MRI of Knee
Presentation
Example – MRI of Knee
Presentation
Presentation

10

10

10
8  
10
 4/10/14  

Image  Types  

RGB   Grey  scale   Binary  

9  
 4/10/14  

Region boundaries
Connected   Components  

To define connected components

we must define regions.

How many white regions are there

in this image?

The number of white regions

actually depends upon the definition
of pixel neighbors.

If white pixels touch only on their sides there are three white
regions. If they touch on their corners and sides then there is one
white region.
7

Neighborhood  
Neighborhoods
Region boundaries
The neighborhood of a pixel is the set
Howof pixels
manyin black
the image that itare
objects touches.
in this image?
The 4-neighbors of p touch p on the
sides (dark pixels). Their coordinates
How
are many white regions?
x 1, y , x 1, y , x, y 1 , x, y 1
The number, once again,
Each pixel
depends onishow
a unitindividual
distance from
(x,y)
pixels connect.

Pixel connection is defined in terms of neighborhoods.

Note: The pixel p is not a member of its neighborhood.
89

10  4

Neighborhoods
 Note: The pixel p is not a member of its neighborhood.
9

4/10/14  

Neighborhoods
Neighborhood  

The d-neighbors of p touch p on

the corners (shaded pixels), their
coordinates are
x 1, y 1 , x 1, y 1 , x 1, y 1 , x 1, y 1

10

5
Neighborhoods
Neighborhood  

The 8-neighbors of p touch p on

either the sides or the corners.

The 8-neighbor set is the union of

the 4-neighbor and d-neighbor set.

Pixels outside the image are not in

the neighborhood.

11

11  

Connectivity
 13

4/10/14  

Connectivity
Connec=vity  

Are the pixels a, b, c, d, e, f, g

connected?

If they are connected then we can say

there is a path from a to g.

14

Connec=vity  
Connected components
A set S of pixels is a connected
component if there is at least one path
in S that joins every pair {p,q} of pixels ! #
in S. The path must contain only pixels $
in S:

• A, B, and C are connected "

components under 4-connectivity.

• B and C are joined under 8- or m-

connectivity.

• Depending on kind of connectivity,

how many holes are in D?
17

Connected components
12  
 4/10/14  

Connec=vity  Example  

Morphological  Processing  
Erosion  

13  
 4/10/14  

Morphological  Processing  
Dila=on  

Connected  Components  
A[er  erosion  

14  
 4/10/14  

Erosion,  dila=on,  open,  close  

Erosion   Dila=on   Open   Close  

Skeletoniza=on  

15  
 4/10/14  

Point processing
Used to transform an image by operating on individual pixels.
Point  
Assume A is an input imagePand
rocessing  
B is the output array produced
by the transformation T:

B x, y T A x, y

Gamma  Correc=on  
•  Can  be  used  to  change  the  appearance  of  an  image  (Gamma  
correc=on  –  Brightness,      s    =      r  γ    )  

16  
 4/10/14  

Histogram  Equaliza=on  

17