Digital Image Processing

Intensity Transformations
and
Spatial Filtering
Background

 Spatial domain process

g ( x, y )  T [ f ( x, y )]

 where f ( x, y ) is the input image, g ( x, y )

is the processed image, and T is an
operator on f, defined over some
neighborhood of ( x, y )
 Neighborhood about a point
 Gray-level transformation function
s  T (r )
 where r is the gray level of f ( x, y ) and
s is the gray level of g ( x, y ) at any
point ( x, y )
 Contrast enhancement
 For example, a thresholding function
 Masks (filters, kernels, templates,
windows)
 A small 2-D array in which the values
of the mask coefficients determine the
nature of the process
Some Basic Gray Level
Transformations
 Image negatives
s  L 1 r
 Enhance white or gray details
 Log transformations
s  c log(1  r )
 Compress the dynamic range of images
with large variations in pixel values
 From the range 0-1.5  106 to the range
0 to 6.2
 Power-law transformations
 
 s  cr or s  c(r   )
   1 maps a narrow range of dark
input values into a wider range of
output values, while   1 maps a
narrow range of bright input values into
a wider range of output values
  : gamma, gamma correction
 Monitor,   2.5
 Piecewise-linear transformation
functions
 The form of piecewise functions can be
arbitrarily complex
 Contrast stretching
 Gray-level slicing
 Bit-plane slicing
Image Enhancement: Histogram
Based Methods
 What is the histogram of a digital image?

The histogram of a digital image with gray values r0 , r1 ,, rL 1

is the discrete function h(rk )  nk
nk
Normalized histogram p (rk ) 
n
nk: Number of pixels with gray value rk

n : total Number of pixels in the image

rk : kth gray level of a pixel

The function p(rk) represents the fraction of the total number

of pixels with gray value rk.
Histogram provides a global description of the appearance of
the image.

If we consider the gray values in the image as realizations of a

random variable R, with some probability density, histogram
provides an approximation to this probability density. In other
words,

Pr( R  rk )  p (rk )
 Some Typical Histograms
The shape of a histogram provides useful information for
contrast enhancement.

Dark image
Bright image

Low contrast image

High contrast image
 Histogram Examples (cont…)

A selection of images and

their histograms
Notice the relationships
between the images and
their histograms
Note that the high contrast
image has the most
evenly spaced histogram
 Histogram Equalization

Spreading out the frequencies in an image (or equalising the image) is a

simple way to improve dark or washed out images.

The histogram equalization is an approach to enhance a given image. The

approach is to design a transformation T(.) such that the gray values in the
output is uniformly distributed in [0, 1].
Let us assume for the moment that the input image to be
enhanced has continuous gray values, with r = 0 representing
black and r = 1 representing white.

We need to design a gray value transformation s = T(r), based

on the histogram of the input image, which will enhance the
image.
 Assume that:

(1) T(r) is a monotonically increasing function for

0 r 1 (preserves order from black to white).

(2) T(r) maps [0,1] into [0,1] (preserves the range of

allowed Gray values).
 Histogram equalization
s  T (r ), 0  r  1
r  T 1 ( s ), 0  s  1
Let us denote the inverse transformation by r T -1(s) . We
assume that the inverse transformation also satisfies the above
two conditions.

We consider the gray values in the input image and output
image as random variables in the interval [0, 1].

Let pin(r) and pout(s) denote the probability density of the

Gray values in the input and output images.
If pin(r) and T(r) are known, and r T -1(s) satisfies
condition 1, we can write (result from probability theory):

 dr 
pout ( s )   pin (r ) 
 ds  r T 1 ( s )

One way to enhance the image is to design a transformation

T(.) such that the gray values in the output is uniformly
distributed in [0, 1], i.e. pout (s) 1, 0  s1

In terms of histograms, the output image will have all

gray values in “equal proportion” .

This technique is called histogram equalization.

Next we derive the gray values in the output is uniformly
distributed in [0, 1].

Consider the transformation

r
s  T (r )   pin ( w)dw, 0  r 1
0

Note that this is the cumulative distribution function (CDF)

of pin (r) and satisfies the previous two conditions.

From the previous equation and using the fundamental

theorem of calculus,

ds
 pin (r )
dr
Therefore, the output histogram is given by

 1 
pout ( s )   pin (r )    1r T 1 ( s )  1, 0  s 1
 pin (r )  r T 1 ( s )

The output probability density function is uniform,

regardless of the input.

Thus, using a transformation function equal to the CDF of

input gray values r, we can obtain an image with uniform gray
values.

This usually results in an enhanced image, with an increase

in the dynamic range of pixel values.
How to implement histogram equalization?

Step 1:For images with discrete gray values, compute:

nk
pin (rk )  0  rk  1 0  k  L 1
n
L: Total number of gray levels
nk: Number of pixels with gray value rk
n: Total number of pixels in the image

Step 2: Based on CDF, compute the discrete version of the

previous transformation :
k
sk  T (rk )   pin (r j ) 0  k  L 1
j 0
  Probability density functions (PDF)
dr
ps ( s )  pr ( r )
ds
r
s  T (r )  ( L  1)  pr ( w)dw
0

ds dT (r ) d  r   ( L  1) p (r )
dr

dr
 ( L  1)

dr  0
p r ( w) dw

 r

1
ps ( s ) 
L 1
 k k nj
sk  T (rk )  ( L  1) pr (rj ) ( L  1) , k  0,1,2,..., L  1
j 0 j 0 n
 Histogram matching (specification)
r
s  T (r )  ( L  1)  pr ( w)dw
0
z
G ( z )  ( L  1)  p z (t )dt  s
0

z  G 1 ( s )  G 1[T (r )]

p z (z ) is the desired PDF

 k k nj
sk  T (rk )  ( L  1) pr (rj )  ( L  1) , k  0,1,2,..., L  1
j 0 j 0 n
k
vk  G ( zk )  ( L  1) p z ( zi ) sk , k  0,1,2,..., L  1
i 0

1
z k  G [T (rk )], k  0,1,2,..., L  1
 Histogram matching
 Obtain the histogram of the given
image, T(r)
 Precompute a mapped level sk for each
level rk
 Obtain the transformation function G
from the given p z (z )
 Precompute z k for each value of sk
 Map rk to its corresponding level sk ;
then map level sk into the final level z k
 Local enhancement
 Histogram using a local neighborhood,
for example 7*7 neighborhood
 Histogram using a local 3*3
neighborhood
 Use of histogram statistics for
image enhancement
 r denotes a discrete random variable
 p (ri ) denotes the normalized
histogram component corresponding to
the ith value of r
 Mean L 1
m   ri p(ri )
i 0
 The nth moment
L 1
 n (r )   (ri  m) p(ri )
n

i 0

 The second moment

L 1
 2 (r )   (ri  m) p(ri )
2

i 0
 Global enhancement: The global mean
and variance are measured over an
entire image
 Local enhancement: The local mean
and variance are used as the basis for
making changes
 rs ,t is the gray level at coordinates
(s,t) in the neighborhood
 p(rs ,t ) is the neighborhood normalized
histogram component
 mean:
mS xy  r
( s ,t )S xy
s ,t p(rs ,t )
 local variance
 2
S xy   [r
( s ,t )S xy
s ,t  mS xy ] p(rs ,t )
2
  E , k0 , k1 , k 2 are specified parameters
 M G is the global mean
 DG is the global standard deviation
 Mapping

 if mS xy  k0 M G
 E  f ( x, y )
g ( x, y )   and k1 DG   S xy  k 2 DG
 f ( x, y ) otherwise

Fundamentals of Spatial Filtering

 The Mechanics of Spatial Filtering

R  w(1,1) f ( x  1, y  1) 
w(1,0) f ( x  1, y )   
w(0,0) f ( x, y )   
w(1,0) f ( x  1, y ) 
w(1,1) f ( x  1, y  1)
 Image size: M N
 Mask size: mn
a b
g ( x, y )    w(s, t ) f ( x  s, y  t )
s   at   b

 a  (m  1) / 2 and b  (n  1) / 2
 x  0,1,2,..., M  1 and y  0,1,2,..., N  1
 Spatial Correlation and Convolution
 Vector Representation of Linear
Filtering
R  w1 z1  w2 z 2  ...  w9 z9
9
  wi zi
i 1
Smoothing Spatial Filters

 Smoothing Linear Filters

 Noise reduction
 Smoothing of false contours
 Reduction of irrelevant detail
 1 9
R   zi
9 i 1
 a b

  w(s, t ) f ( x  s, y  t )
g ( x, y )  s   at   b
a b

  w(s, t )
s   at   b
 Order-statistic filters
 median filter: Replace the value of a
pixel by the median of the gray levels
in the neighborhood of that pixel
 Noise-reduction
Sharpening Spatial Filters

 Foundation
 The first-order derivative
f
 f ( x  1)  f ( x)
x
 The second-order derivative

 f
2
 f ( x  1)  f ( x  1)  2 f ( x)
x 2
 Use of second derivatives for
enhancement-The Laplacian
 Development of the method

 2
f  2
f
 f  2  2
2

x y
 f
2
 f ( x  1, y )  f ( x  1, y )  2 f ( x, y )
x 2

 f
2
 f ( x, y  1)  f ( x, y  1)  2 f ( x, y )
y 2
  f  [ f ( x  1, y )  f ( x  1, y )  f ( x, y  1) 
2

f ( x, y  1)]  4 f ( x, y )
 if the center coefficient
 f ( x, y )   2 f ( x, y ) of the Laplacian mask

 is negative
g ( x, y )  
if the center coefficient

 f ( x, y )   f ( x, y ) of the Laplacian mask
2


 is positive
  Simplifications

g ( x, y )  f ( x, y )  [ f ( x  1, y )  f ( x  1, y )  f ( x, y  1) 
f ( x, y  1)]  4 f ( x, y )
 5 f ( x, y )  [ f ( x  1, y )  f ( x  1, y )  f ( x, y  1) 
f ( x, y  1)]
 Unsharp masking and highboost
filtering
 Unsharp masking
 Substract a blurred version of an image
from the image itself

g mask ( x, y )  f ( x, y )  f ( x, y )

 f ( x, y ) : The image, f ( x, y ) : The

blurred image
g ( x, y )  f ( x, y )  k * g mask ( x, y ) ,k 1
  High-boost filtering

g ( x, y )  f ( x, y )  k * g mask ( x, y ) ,k 1
 Using first-order derivatives for
(nonlinear) image sharpening—The
gradient
 f 
Gx   x 
f      f 
G y   
 y 
 The magnitude is rotation invariant
(isotropic)


f  mag(f )  G  G 2
x
2
y 
1
2

1
 f  2  f  2  2

      
 x   y  

f  G x  G y
 Computing using cross differences,
Roberts cross-gradient operators
G x  ( z 9  z5 ) and G y  ( z8  z 6 )


f  ( z9  z5 )  ( z8  z6 )
2 2

1
2

f  z9  z5  z8  z6
 Sobel operators
 A weight value of 2 is to achieve some
smoothing by giving more importance to
the center point

f  ( z7  2 z8  z9 )  ( z1  2 z 2  z3 )
 ( z3  2 z6  z9 )  ( z1  2 z 4  z7 )
Combining Spatial Enhancement
Methods

 An example
 Laplacian to highlight fine detail
 Gradient to enhance prominent edges
 Smoothed version of the gradient
image used to mask the Laplacian
image
 Increase the dynamic range of the gray
levels by using a gray-level
transformation