CS589-04 Digital Image Processing

Lecture 4. Image Restoration and

Reconstruction

Spring 2008
New Mexico Tech
 Image Restoration

► Image restoration: recover an image that has been

degraded by using a prior knowledge of the degradation
phenomenon.

► Model the degradation and applying the inverse process in

order to recover the original image.

03/21/22 2
 A Model of Image Degradation/Restoration
Process

► Degradation
 Degradation function H
 Additive noise  ( x, y )

03/21/22 3
 A Model of Image Degradation/Restoration
Process

If H is a linear, position-invariant process, then

the degraded image is given in the spatial domain by
g ( x, y )  h( x, y ) f ( x, y )   ( x, y )

03/21/22 4
 A Model of Image Degradation/Restoration
Process

The model of the degraded image is given in

the frequency domain by

G (u, v)  H (u, v) F (u, v)  N (u, v)

03/21/22 5
 Noise Sources

► The principal sources of noise in digital images arise during

image acquisition and/or transmission

 Image acquisition
e.g., light levels, sensor temperature, etc.

 Transmission
e.g., lightning or other atmospheric disturbance in wireless
network

03/21/22 6
 Noise Models (1)

► White noise
 The Fourier spectrum of noise is constant

► With the exception of spatially periodic noise, we assume

 Noise is independent of spatial coordinates
 Noise is uncorrelated with respect to the image itself

03/21/22 7
 Noise Models (2)

 Gaussian noise
Electronic circuit noise, sensor noise due to poor illumination and/or
high temperature

 Rayleigh noise
Range imaging

03/21/22 8
 Range Imaging (1)

 Short for High Dynamic Range Imaging. HDRI is an imaging

technique that allows for a greater dynamic range of exposure
than would be obtained through any normal imaging process.

 It is now popularly used to refer to the process of

tone mapping* together with bracketed** exposures of normal
digital images, giving the end result a high, often exaggerated
dynamic range

* Tone mapping is a technique used in image processing and computer graphics to map
a set of colours to another; often to approximate the appearance of HDRI in media with a
more limited dynamic range
** bracketing is the general technique of taking several shots of the same subject using
different or the same camera settings

http://en.wikipedia.org/wiki/High_dynamic_range_imaging
http://www.webopedia.com/TERM/H/High_Dynamic_Range_Imaging.html

03/21/22 9
 Range Imaging (2)

 The intention of HDRI is to accurately represent the wide

range of intensity levels found in real scenes ranging from
direct sunlight to shadows

 HDRI, also called HDR (High Dynamic Range) is a feature

commonly found in high-end graphics and imaging software

http://en.wikipedia.org/wiki/High_dynamic_range_imaging
http://www.webopedia.com/TERM/H/High_Dynamic_Range_Imaging.html

03/21/22 10
 Range Imaging: Examples (1)
http://en.wikipedia.org/wiki/High_dynamic_range_imaging

Tower Bridge
in Sacramento,
CA

03/21/22 11
 Sydney Harbour
Bridge HDRi
Range Imaging: Examples (2) produces greater
detail and fewer
http://en.wikipedia.org/wiki/High_dynamic_range_imaging shadows

03/21/22 12
03/21/22 Old Saint Paul’s Wellinton, New Zealand 13
 http://en.wikipedia.org/wiki/Image:HDRI-Example.jpg

03/21/22 14
 Noise Models (3)

 Erlang (gamma) noise: Laser imaging

 Exponential noise: Laser imaging

 Uniform noise: Least descriptive; Basis for numerous random

number generators

 Impulse noise: quick transients,

such as faulty switching

03/21/22 15
 Gaussian Noise (1)

The PDF of Gaussian random variable, z, is given by

1  ( z  z ) 2 /2 2
p( z )  e
2

where, z represents intensity

z is the mean (average) value of z
 is the standard deviation

03/21/22 16
 Gaussian Noise (2)

The PDF of Gaussian random variable, z, is given by

1  ( z  z ) 2 /2 2
p( z )  e
2

 70% of its values will be in the range

(    ), (    )
 95% of its values will be in the range

(   2 ), (  2 )
03/21/22 17
 Rayleigh Noise

The PDF of Rayleigh noise is given by

2  ( z  a )2 / b
 ( z  a )e for z  a
p( z )   b
 0 for z  a

The mean and variance of this density are given by

z  a  b / 4
2 b(4   )
 
4

03/21/22 18
 Erlang (Gamma) Noise

The PDF of Erlang noise is given by

 a b z b 1  az
 e for z  0
p ( z )   (b  1)!
0 for z  a


The mean and variance of this density are given by

z b/a
 2  b / a2

03/21/22 19
 Exponential Noise

The PDF of exponential noise is given by

 ae  az for z  0
p( z )  
0 for z  a

The mean and variance of this density are given by

z  1/ a
 2  1/ a 2

03/21/22 20
 Uniform Noise

The PDF of uniform noise is given by

 1
 for a  z  b
p( z )   b  a
 0 otherwise

The mean and variance of this density are given by

z  ( a  b) / 2
 2  (b  a) 2 /12

03/21/22 21
 Impulse (Salt-and-Pepper) Noise

The PDF of (bipolar) impulse noise is given by

 Pa for z  a

p( z )   Pb for z  b
0 otherwise


if b  a, gray-level b will appear as a light dot,

while level a will appear like a dark dot.

If either Pa or Pb is zero, the impulse noise is called

unipolar
03/21/22 22
03/21/22 23
 Examples of Noise: Original Image

03/21/22 24
 Examples of Noise: Noisy Images(1)

03/21/22 25
 Examples of Noise: Noisy Images(2)

03/21/22 26
 Periodic Noise

► Periodic noise in an image arises typically from electrical or

electromechanical interference during image acquisition.

► It is a type of spatially dependent noise

► Periodic noise can be reduced significantly via frequency

domain filtering

03/21/22 27
 An Example of Periodic Noise

03/21/22 28
 Estimation of Noise Parameters (1)
The shape of the histogram identifies the closest PDF match

03/21/22 29
 Estimation of Noise Parameters (2)

Consider a subimage denoted by S , and let ps ( zi ), i  0, 1, ..., L -1,

denote the probability estimates of the intensities of the pixels in S .
The mean and variance of the pixels in S:
L 1
z   z i p s ( zi )
i 0
L 1
and  2   ( zi  z ) 2 p s ( z i )
i 0

03/21/22 30
 Restoration in the Presence of Noise Only
̶ Spatial Filtering

Noise model without degradation

g ( x, y )  f ( x, y)   ( x, y)
and
G (u, v)  F (u, v)  N (u, v)

03/21/22 31
 Spatial Filtering: Mean Filters (1)

Let S xy represent the set of coordinates in a rectangle

subimage window of size m  n, centered at ( x, y ).

Arithmetic mean filter

f ( x, y )  1

mn ( s ,t )S xy
g ( s, t )

03/21/22 32
 Spatial Filtering: Mean Filters (2)

Geometric mean filter

1
  mn
f ( x, y )  
 g ( s, t ) 
 ( s ,t )S xy 

Generally, a geometric mean filter achieves smoothing

comparable to the arithmetic mean filter, but it tends to
lose less image detail in the process

03/21/22 33
 Spatial Filtering: Mean Filters (3)
Harmonic mean filter
f ( x, y )  mn
1

( s ,t )S xy g ( s, t )

It works well for salt noise, but fails for pepper noise.
It does well also with other types of noise like Gaussian
noise.

03/21/22 34
 Spatial Filtering: Mean Filters (4)

Contraharmonic mean filter

 g ( s, t )Q 1
f ( x, y )  ( s ,t )S xy


( s ,t )S xy
g ( s, t )Q

Q is the order of the filter.

It is well suited for reducing the effects of salt-and-

pepper noise. Q>0 for pepper noise and Q<0 for salt
noise.
03/21/22 35
 Spatial Filtering: Example (1)

03/21/22 36
 Spatial Filtering: Example (2)

03/21/22 37
 Spatial Filtering: Example (3)

03/21/22 38
 Spatial Filtering: Order-Statistic Filters (1)

Median filter
f ( x, y)  median g ( s, t )
( s ,t )S xy

Max filter
f ( x, y)  max g ( s, t )
( s ,t )S xy

Min filter
f ( x, y)  min g ( s, t )
( s ,t )S xy

03/21/22 39
 Spatial Filtering: Order-Statistic Filters (2)

Midpoint filter
f ( x, y )  1  max g ( s, t )  min g ( s, t )
2  ( s ,t )S xy ( s ,t )S xy 

03/21/22 40
 Spatial Filtering: Order-Statistic Filters (3)

Alpha-trimmed mean filter

f ( x, y )  1

mn  d ( s ,t )S xy
g r ( s, t )

We delete the d / 2 lowest and the d / 2 highest intensity values of

g ( s, t ) in the neighborhood S xy . Let g r ( s, t ) represent the remaining
mn - d pixels.

03/21/22 41
03/21/22 42
03/21/22 43
03/21/22 44
 Spatial Filtering: Adaptive Filters (1)

Adaptive filters

The behavior changes based on statistical characteristics

of the image inside the filter region defined by the mхn
rectangular window.

The performance is superior to that of the filters

discussed

03/21/22 45
 Adaptive Filters:
Adaptive, Local Noise Reduction Filters (1)

S xy : local region
The response of the filter at the center point (x,y) of S xy
is based on four quantities:
(a) g ( x, y ), the value of the noisy image at (x, y );
(b)  2 , the variance of the noise corrupting f ( x, y )
to form g ( x, y);
(c) mL , the local mean of the pixels in S xy ;
(d)  L2 , the local variance of the pixels in S xy .

03/21/22 46
 Adaptive Filters:
Adaptive, Local Noise Reduction Filters (2)

The behavior of the filter:

(a) if  2 is zero, the filter should return simply the value
of g ( x, y ).
(b) if the local variance is high relative to  2 , the filter
should return a value close to g ( x, y );
(c) if the two variances are equal, the filter returns the
arithmetic mean value of the pixels in S xy .

03/21/22 47
 Adaptive Filters:
Adaptive, Local Noise Reduction Filters (3)

An adaptive expression for obtaining f ( x, y)

based on the assumptions:
2
f ( x, y )  g ( x, y )   
2  g ( x, y )  m L 
 L

03/21/22 48
03/21/22 49
 Adaptive Filters:
Adaptive Median Filters (1)

The notation:
zmin  minimum intensity value in S xy
zmax  maximum intensity value in S xy
zmed  median intensity value in S xy
z xy  intensity value at coordinates ( x, y)
S max  maximum allowed size of S xy
03/21/22 50
 Adaptive Filters:
Adaptive Median Filters (2)
The adaptive median-filtering works in two stages:
Stage A:
A1 = zmed  zmin ; A2 = zmed  zmax
if A1>0 and A2<0, go to stage B
Else increase the window size
if window size  S max , repeat stage A; Else output zmed
Stage B:
B1 = z xy  zmin ; B2 = z xy  zmax
if B1>0 and B2<0, output z xy ; Else output zmed
03/21/22 51
 Adaptive Filters:
Adaptive Median Filters (2)
The adaptive median-filtering works in two stages:
Stage A: The median filter
A1 = zmed  zmin ; A2 = zmed  zmaxoutput is an impulse
or not
if A1>0 and A2<0, go to stage B
Else increase the window size
if window size  S max , repeat stage A; Else output zmed
Stage B:
The processed point
B1 = z xy  zmin ; B2 = z xy  zmax is an impulse or not
if B1>0 and B2<0, output z xy ; Else output zmed
03/21/22 52
 Example:
Adaptive Median Filters

03/21/22 53
 Periodic Noise Reduction by Frequency
Domain Filtering

The basic idea

Periodic noise appears as concentrated bursts of energy

in the Fourier transform, at locations corresponding to
the frequencies of the periodic interference

Approach

A selective filter is used to isolate the noise

03/21/22 54
 Perspective Plots of Bandreject Filters

03/21/22 55
 A Butterworth bandreject
filter of order 4, with the
appropriate radius and
width to enclose
completely the noise
impulses

03/21/22 56
 Perspective Plots of Notch Filters

03/21/22 57
03/21/22 58
 Several interference components are present, the methods discussed
in the preceding sections are not always acceptable because they
remove much image information
The components tend to have broad skirts that carry information
about the interference pattern and the skirts are not always easily
detectable.
03/21/22 59
 Optimum Notch Filtering

It minimizes local variances of the restored estimated

f ( x, y )

Procedure for restoration tasks in multiple periodic

interference

Isolate the principal contributions of the interference

pattern

Subtract a variable, weighted portion of the pattern

from the corrupted image
03/21/22 60
 Optimum Notch Filtering: Step 1

Extract the principal frequency components of

the interference pattern

Place a notch pass filter at the location of each spike.

N (u , v)  H NP (u , v)G (u, v)
 ( x, y )  1 H NP (u, v)G (u, v)

03/21/22 61
 Optimum Notch Filtering: Step 2 (1)

Filtering procedure usually yields only an approximation of the

true pattern. The effect of components not present in the estimate
of  (x, y ) can be minimized instead by subtracting from g ( x, y )
a weighted portion of  (x, y ) to obtain an estimate of f ( x, y ):

f ( x, y )  g ( x, y )  w( x, y ) ( x, y )

One approach is to select w( x, y ) so that the variance of the estimate f ( x, y)

is minimized over a specified neighborhood of every point ( x, y ).
03/21/22 62
 Optimum Notch Filtering: Step 2 (2)

The local variance of f ( x, y ):

2
1 a b
 f ( x  s, y  t )  f ( x, y ) 
2
 ( x, y )   
(2a  1)(2b  1) s  a t b  

03/21/22 63
Assume that w(x,y) remains
essentially constant over the
Optimum
neighborhood gives theNotch Filtering: Step (3)
approximation
w(x+s,
The localy+t) of f ( x, y ):
= w(x,y)
variance
2
1 a b
 f ( x  s, y  t )  f ( x, y ) 
2
 ( x, y )   
(2a  1)(2b  1) s  a t  b  

2
b  g ( x  s , y  t )  w( x  s, y  t ) ( x  s, y  s ) 
1 a
 
   
(2a  1)(2b  1) s  a t  b   g ( x, y )  w( x, y ) ( x, y ) 

   

2
b  g ( x  s , y  t )  w( x, y ) ( x  s , y  s ) 
1 a
 
   

(2a  1)(2b  1) s  a t  b  g ( x, y )  w( x, y ) ( x, y ) 

   
03/21/22 64
 Optimum Notch Filtering: Step (4)

The local variance of f ( x, y):

2
1 a b  g ( x  s, y  t )  w( x, y ) ( x  s, y  s ) 
2
 ( x, y )   
(2a  1)(2b  1) s  a t  b   g ( x, y )  w( x, y ) ( x, y ) 

   

2
2  ( x, y )
To minimize  ( x, y ) , 0
w( x, y )
for w( x, y ), the result is
g ( x, y ) ( x, y )  g ( x, y ) ( x, y )
w( x, y )  2
2
 ( x, y )   ( x, y )
03/21/22 65
 Optimum Notch Filtering: Example

03/21/22 66
 Optimum Notch Filtering: Example

03/21/22 67
 Optimum Notch Filtering: Example

03/21/22 68
 Optimum Notch Filtering: Example

03/21/22 69
 Linear, Position-Invariant Degradations

g ( x, y )  H  f ( x , y )    ( x , y )

03/21/22 70
 Linear, Position-Invariant Degradations

H is linear
H af1 ( x, y )  bf 2 ( x, y )   aH  f1 ( x, y )   bH  f 2 ( x, y ) 
f1 and f 2 are any two input images.

An operator having the input-output relationship

g ( x, y )  H  f ( x, y )  is said to be position invariant
if
H  f ( x   , y   )  g ( x   , y   )
for any f ( x, y ) and any  and  .
03/21/22 71
 Linear, Position-Invariant Degradations
 
f ( x, y )    f ( ,  ) ( x   , y   ) d d 
 
Assume for a moment that  ( x, y )  0
if H is a linear operator,
g ( x, y )  H  f ( x, y ) 
 
 H   f ( ,  ) ( x   , y   ) d d  
 Superposition (or
    Fredholm)
  integral of the
  H  f ( ,  ) ( x   , y   )  d d  first kind
 
 
  f ( ,  ) H  ( x   , y   )  d d  Impulse
 
response
03/21/22 72
 Linear, Position-Invariant Degradations

Assume for a moment that  ( x, y )  0

if H is a linear operator and position invariant,
H  ( x   , y   )   h( x   , y   )
g ( x, y )  H  f ( x, y ) 
 
  f ( ,  ) H  ( x   , y   )  d d 
  Convolution
  integral in 2-D
  f ( ,  )h( x   , y   )d d 
 

03/21/22 73
 Linear, Position-Invariant Degradations

In the presence of additive noise,

if H is a linear operator and position invariant,

 
g ( x, y )    f ( ,  ) h( x   , y   )d d    ( x, y )
 

 h ( x, y ) f ( x, y )   ( x , y )

G (u, v)  H (u, v) F (u, v)  N (u , v)

03/21/22 74
 Estimating the Degradation Function

► Three principal ways to estimate the degradation function

1. Observation

2. Experimentation

3. Mathematical Modeling

03/21/22 75
 Mathematical Modeling (1)

► Environmental conditions cause degradation

A model about atmospheric turbulence

 k ( u 2  v 2 )5/6
H (u , v )  e
k : a constant that depends on
the nature of the turbulence

03/21/22 76
03/21/22 77
 Mathematical Modeling (2)

► Derive a mathematical model from basic principles

E.g., An image blurred by uniform linear motion between

the image and the sensor during image acquisition

03/21/22 78
 Mathematical Modeling (3)

Suppose that an image f ( x, y ) undergoes planar motion,

x0 (t ) and y0 (t ) are the time-varying components of motion
in the x- and y -directions, respectively.
The optical imaging process is perfect. T is the duration
of the exposure. The blurred image g ( x, y )
T
g ( x, y )   f  x  x0 (t ), y  y0 (t ) dt
0

03/21/22 79
 Mathematical Modeling (4)
T
g ( x, y )   f  x  x0 (t ), y  y0 (t ) dt
0
 
G (u, v)    g ( x, y ) e  j 2 ( ux  vy )
dxdy
 
 T 
    f  x  x0 (t ), y  y0 (t ) dt  e  j 2 ( ux  vy ) dxdy

  
 0 
T
    f  x  x (t ), y  y (t )  e  j 2 ( ux  vy ) dxdy dt
0    0 0 
T  j 2 ux0 ( t )  vy0 ( t ) 
  F (u , v)e dt
0
T  j 2 ux0 ( t )  vy0 ( t ) 
 F (u , v )  e dt
0
03/21/22 80
 Mathematical Modeling (4)

T  j 2 ux0 ( t )  vy0 ( t )
H (u , v )   e dt
0

Suppose that the image undergoes uniform linear motion

in the x-direction only, at a rate given by x0 (t )  at / T .
T
H (u, v)   e  j 2 ux0 ( t ) dt
0
T
  e  j 2 uat /T dt
0

T  j ua
 sin( ua)e
 ua
03/21/22 81
 Mathematical Modeling (5)

Suppose that the image undergoes uniform linear motion

in the x-direction and y -direction, at a rate given by
x0 (t )  at / T and y0 (t )  bt / T
T  j 2 ux0 ( t )  vy0 ( t )
H (u , v )   e dt
0
T
 e  j 2 [ ua  vb ]t /T
dt
0

T
 sin  (ua  vb)  e  j (ua vb )
 (ua  vb)
03/21/22 82
03/21/22 83
 Inverse Filtering

An estimate of the transform of the original image

 G (u, v)
F (u , v ) 
H (u, v)

 F (u , v) H (u , v)  N (u, v)
F (u , v) 
H (u , v)
N (u , v)
 F (u, v) 
H (u , v)
03/21/22 84
 Inverse Filtering

 N (u , v)
F (u, v)  F (u , v) 
H (u , v)

1. We can't exactly recover the undegraded image

because N (u, v) is not known.
2. If the degradation function has zero or very
small values, then the ratio N (u, v) / H (u, v) could
easily dominate the estimate F  (u, v).

03/21/22 85
 Inverse Filtering

EXAMPLE

The image in Fig. 5.25(b) was inverse filtered using the

exact inverse of the degradation function that generated
that image. That is, the degradation function is
2 5/6
 k u  M /2   ( v  N / 2)2 
H (u , v )  e  
, k  0.0025

03/21/22 86
 Inverse Filtering

One approach is to limit the filter frequencies to values near the origin.

EXAMPLE

The image in Fig. 5.25(b) was inverse filtered using the

exact inverse of the degradation function that generated
that image. That is, the degradation function is
2 5/6
 k u  M /2   ( v  N /2) 2 
H (u , v )  e  

k  0.0025, M  N  480.
03/21/22 87
 The poor performance
of direct inverse
filtering in general

A Butterworth
lowpass
function of
order 10

03/21/22 88
 Minimum Mean Square Error (Wiener)
Filtering
 N. Wiener (1942)

 Objective
Find an estimate of the uncorrupted image such that the mean
square error between them is minimized


e 2  E ( f  f ) 2 

03/21/22 89
 Minimum Mean Square Error (Wiener)
Filtering

The minimum of the error function is given in the frequency domain

by the expression

  H *(u, v) S f (u, v) 
F (u, v)   2  G (u, v)
 S f (u , v) | H (u , v) |  S (u , v) 
 H *(u, v) 
 2  G (u , v)
 | H (u , v) |  S (u , v ) / S f (u, v) 
 1 | H (u, v) |2 
 2  G (u, v)
 H (u, v) | H (u , v) |  S (u, v) / S f (u, v) 
03/21/22 90
 Minimum Mean Square Error (Wiener)
Filtering

  1 | H (u , v ) |2 
F (u , v)   2  G (u, v)
 H (u , v) | H (u, v) |  S (u, v) / S f (u, v) 

H (u, v) : degradation function

H *(u , v): complex conjugate of H (u, v)
| H (u , v) |2  H *(u, v) H (u, v)
S (u, v) | N (u , v) |2  power spectrum of the noise
S f (u , v) | F (u , v) |2  power spectrum of the undegraded image

03/21/22 91
 Minimum Mean Square Error (Wiener)
Filtering

2
  1 | H (u , v ) | 
F (u , v)   2  G (u, v)
 H (u , v) | H (u, v) |  K 

K is a specified constant. Generally, the value of K

is chosen interactively to yield the best visual results.

03/21/22 92
 Minimum Mean Square Error (Wiener)
Filtering

03/21/22 93
Left:
degradated
image

Middle:
inverse
filtering

Right:
Wiener
filtering

03/21/22 94
 Some Measures (1)

Singal-to-Noise Ratio (SNR)

M 1 N 1

 | F (u , v ) | 2

SNR  u 0 v 0
M 1 N 1

 | N (u
u 0 v 0
, v ) |2

This ratio gives a measure of the level of information

bearing singal power to the level of noise power.

03/21/22 95
 Some Measures (2)

Mean Square Error (MSE)

2
1 M 1 N 1
MSE=
MN
   f ( x, y)  f ( x, y) 
x 0 y 0

Root-Mean-Sqaure-Error (RMSE)
M 1 N 1
f ( x, y ) 2

RMSE  M 1 N 1
u 0 v 0

f ( x, y ) |2
 | f (
u 0 v 0
x , y ) 

03/21/22 96
 Constrained Least Squares Filtering

► In Wiener filter, the power spectra of the undegraded

image and noise must be known. Although a constant
estimate is sometimes useful, it is not always suitable.

► Constrained least squares filtering just requires the mean

and variance of the noise.

03/21/22 97
 Constrained Least Squares Filtering

 H *(u, v) 
F (u, v)   2 2
G (u , v )
 | H (u , v) |  | P (u, v) | 

P(u , v) is the Fourier transform of the function

 0 1 0 
p( x, y )   1 4 1
 0 1 0 
 is a parameter

03/21/22 98
 Examples

03/21/22 99
 Geometric Mean Filter

1

 H *(u, v )   | H (u, v) | 2 

F (u , v)     G (u, v)
2
 | H (u , v ) | 
2
 | H (u, v) |    S (u, v) / S f (u, v)  

  1: inverse filter
 =0: parametric Wiener filter
 =1/2: geometric mean filter

03/21/22 100
 Image Reconstruction from Projection

► Reconstruct an image from a series of projections

X-ray computed tomography (CT)

“Computed tomography is a medical imaging method employing

tomography where digital geometry processing is used to generate a
three-dimensional image of the internals of an object from a large
series of two-dimensional X-ray images taken around a single axis of
rotation.”

http://en.wikipedia.org/wiki/Computed_tomography

03/21/22 101
 Backprojection

“ In computed tomography or other imaging techniques

requiring reconstruction from multiple projections, an
algorithm for calculating the contribution of each voxel of
the structure to the measured ray data, to generate an
image; the oldest and simplest method of image
reconstruction. “

http://www.medilexicon.com/medicaldictionary.php?t=9165

03/21/22 102
 Image Reconstruction: Introduction
Intensity is
Soft, uniform proportional to
tissue absorption

Uniform with
higher absorption
Tumor

03/21/22 103
 Image Reconstruction: Introduction

03/21/22 104
03/21/22 105
03/21/22 106
 Other CTs

► Electron beam CT (Fifth-generation CT)

Electron beam tomography (EBCT) was introduced in the early 1980s, by
medical physicist Andrew Castagnini, as a method of improving the temporal
resolution of CT scanners.
High cost of EBCT equipment, and poor flexibility

► Helical (or spiral) cone beam computed tomography (sixth-generation)

A type of three dimensional computed tomography (CT) in which the source
(usually of x-rays) describes a helical trajectory relative to the object while a
two dimensional array of detectors measures the transmitted radiation on part
of a cone of rays emitting from the source

http://en.wikipedia.org/wiki/Computed_tomography

03/21/22 107
 Other CTs

► Multislice CT (seventh-generation)

► The major benefit of multi-slice CT

 Significant increase in detail

 Utilizes X-ray tubes more economically
 Reducing cost and potentially reducing dosage

03/21/22 108
 Projections and the Radon Transform

03/21/22 109
 Projections and the Radon Transform

 
g (  j , k )    f ( x, y ) ( x cos  k  y sin  k   j )dxdy
 
03/21/22 110
 Projections and the Radon Transform

► Radon transform gives the projection (line integral) of

f(x,y) along an arbitrary line in the xy-plane

 
  f   g (  , )    f ( x, y ) ( x cos   y sin    )dxdy
 

M 1 N 1
  f   g (  ,  )    f ( x, y ) ( x cos   y sin    )
x 0 y 0

03/21/22 111
Example: Using the Radon transform to obtain
the projection of a circular region

► Assume that the circle is centered on the origin of the xy-plane.

Because the object is circularly symmetric, its projections are the

same for all angles, so we just check the projection for   0

A x2  y2  r 2
f ( x, y )  
0 otherwise

03/21/22 112
Example: Using the Radon transform to obtain
the projection of a circular region
 
g (  , )    f ( x, y ) ( x cos   y sin    )dxdy
 
 
  f ( x, y ) ( x   )dxdy
 

 f (  , y )dy


r2  2
 f (  , y )dy
 r2 2

r2  2
 2 2
Ady
 r 

2 A r 2   2 | |  r

 0 otherwise
03/21/22 113
 2 A r 2   2 | |  r
g ( )  
 0 otherwise

03/21/22 114
 Sinogram: The Result of Radon Transform

► Sinogram: the
result of Radon
transform is
displayed as an
image with
 and  as
rectilinear
coordinates

03/21/22 115
 Image Reconstruction

f ( x, y )  g ( x cos   y sin  ,  )

f ( x, y )   f ( x, y )d
0

f ( x, y )   f ( x, y )
 0

A back-projected image formed is referred to as a laminogram

03/21/22 116
 Examples: Laminogram

03/21/22 117
 The Fourier-Slice Theorem

For a given value of  , the 1-D Fourier transform

of a projection with respect to  is

G ( w,  )   g (  ,  )e  j 2 d 


  
G ( ,  )     f ( x, y ) ( x cos   y sin    )e  j 2 d  dxdy
  
 
f ( x, y )   ( x cos   y sin    )e  j 2 d  dxdy
 
 
    
 
  f ( x, y )e  j 2 ( x cos  y sin  ) dxdy
 

03/21/22 118
 The Fourier-Slice Theorem

G ( w,  )   g (  ,  )e  j 2 d 

 
G ( ,  )    f ( x, y )e  j 2 ( x cos  y sin  ) dxdy
 
 
   f ( x, y )e  j 2 (ux  vy ) dxdy 

    u  w cos ,v  w sin 
  F (u, v) u  w cos ,v  w sin 
 F ( w cos  , w sin  )
Fourier-slice theorem: The Fourier tansform of a projection is a slice of
the 2-D Fourier transform of the region from which the projection was
obtained
03/21/22 119
 Illustration of the Fourier-slice theorem

03/21/22 120
 Reconstruction Using Parallel-Beam Filtered
Backprojections
 
f ( x, y )    F (u , v)e j 2 ( ux  vy ) dudv
 

Let u  w cos  , v  w sin  , then dudv  wdwd ,

2 
f ( x, y )    F ( w cos  , w sin  )e j 2 w( x cos  y sin  ) wdwd
0 0
2 
  G ( w,  )e j 2 w( x cos  y sin  ) wdwd
0 0

G ( w,   180 )  G (  w,  )
 
f ( x, y )    | w | G ( w,  )e j 2 w( x cos  y sin  ) dwd
0 
03/21/22 121
 Reconstruction Using Parallel-Beam Filtered
Backprojections

 
f ( x, y )    | w | G ( w,  )e j 2 w ( x cos  y sin  )
dwd
0 

It’s not
   | w | G ( w,  )e j 2 w dw 
integrable 0 
     x cos  y sin 
d

Approach:
Window the ramp so it becomes zero outside of a defined frequency
interval. That is, a window band-limits the ramp filter.

03/21/22 122
 Hamming / Hann Widow

 2 w
c  (c  1) cos 0  w  ( M  1)
h( w)   M 1
 0 otherwise

c  0.54, the function is called the Hamming window

c  0.5, the function is called the Han window

03/21/22 123
 The Plot of Hamming Widow

03/21/22 124
 Filtered Backprojection

The complete, filtered backprojection (to obtain the

reconstructed image f(x,y) ) is described as follows:

1. Compute the 1-D Fourier transform of each projection

2. Multiply each Fourier transform by the filter function |w|
which has been multiplied by a suitable (e.g., Hamming)
window
3. Obtain the inverse 1-D Fourier transform of each
resulting filtered transform
4. Integrate (sum) all the 1-D inverse transforms from step
3
03/21/22 125
 Examples: Filtered Backprojection

03/21/22 126
 Examples: Filtered Backprojection

03/21/22 127
Implementation of Filtered Backprojection in
Spatial Domain
► Fourier transform of the product of two frequency domain
functions is equal to the convolution of the spatial
representation
► Let s(p) denote the inverse Fourier transform of |w|

f ( x, y )     | w | G ( w,  )e j 2 w dw
0 
     x cos  y sin 
d

  s(  ) g (  , ) d
0   x cos  y sin 

   g (  ,  ) s( x cos   y sin    ) d  d
0   
03/21/22 128
 Reconstruction Using Fan-Beam Filtered
Backprojections

  

  D sin 

03/21/22 129
 Reconstruction Using Fan-Beam Filtered
Backprojections
Objects are encompassed within a circular area of radius T about
the origin of the plane, or g( , )=0 for | |>T

f ( x, y )     g (  ,  ) s( x cos   y sin    )d  d
0   
1 2 T
   g (  ,  ) s( x cos   y sin    )d d
2 0 T
x  r cos  ; y  r sin 
x cos   y sin   r cos  cos   r sin  sin 
 r cos(   )
03/21/22 130
 Reconstruction Using Fan-Beam Filtered
Backprojections

x  r cos  ; y  r sin 
x cos   y sin   r cos  cos   r sin  sin 
 r cos(   )
1 2 T
f ( x, y )    g (  ,  ) s r cos(   )    d d
2 0 T
      D sin 

d  d  D cos  d d 

03/21/22 131
Reconstruction Using Fan-Beam Filtered
Backprojections

d  d  D cos  d d 
1 2 T
f ( x, y )    g (  ,  ) s r cos(   )    d d
2 0 T
1 2  sin 1 (T / D )
   g ( D sin  ,    ) s r cos(     )  D sin   D cos  d d 
2    sin 1
(  T / D )

03/21/22 132
 Reconstruction Using Fan-Beam Filtered
Backprojections
1 2 T
f ( x, y)    g (  ,  ) s r cos(   )    d d
2 0 T
1 2  sin1 (T / D )
   g ( D sin  ,    ) s r cos(     )  D sin   D cos  d d 
2    sin 1
(  T / D )

1 2  m
f (r ,  )    p ( ,  ) s  R sin  '  D cos  d d 
2 0  m
2
  
s ( R sin  )    s ( )
 R sin  
2 1  m  d
f (r ,  )    
R 2   m
q ( ,  ) h  '  d 
0 
2
1  
h( )    s ( ), q ( ,  )  p( ,  ) D cos 
03/21/22 2  sin   133
03/21/22 134
03/21/22 135