Statistical Operations

• Gray-level transformation
• Histogram equalization
• Multi-image operations

240-373 Image Processing 1

 Histogram

• If the number of pixels at each gray level in a

n image is counted (may use the following co
de fragment)

for (row=0; row<rowmax; roww++)

for (col=0; col=colmax; col++)
{
count[image[row,col]]++;
}

240-373 Image Processing 2

 Histogram

• The array “count” can be plotted to represent

a “histogram” of the image as the number of
pixels at particular gray level
• The histogram can yield useful information
about the nature of the image. An image may
be too bright or too dark.

240-373 Image Processing 3

 Histogram Illustration

Histogram
10
9
1 2 3 2 3 2 8
0 0 1 2 1 4 7
4 4 2 1 2 1 6
1 2 1 2 1 2 5
5 4 2 1 4 0
4
3
2
1
0 0 1 2 3 4 5

240-373 Image Processing 4

 Global Attenuation in Brightness

• To alter the brightness of an image by adding

or subtracting all pixel values with a constant

for (row=0; row<rowmax; roww++)

for (col=0; col=colmax; col++)
{
image[row,col] += constant;
}

240-373 Image Processing 5

 Intensity Transformation
1
Adjusted Image

0.5

0
0 0.5 1

Histogram Histogram

240-373 Image Processing 6

 Intensity Transformation
1
Adjusted Image

0.5

0
0 0.5 1

Histogram Histogram

240-373 Image Processing 7

 Intensity Transformation
1
Adjusted Image

0.5

0
0 0.5 1

Histogram Histogram

240-373 Image Processing 8

 Intensity Transformation
1
Adjusted Image

0.5

0
0 0.5 1

Histogram Histogram

240-373 Image Processing 9

 Thresholding

• Use:
– To remove the gray level trends in an image
– To make gray level more discrete
– To segment or split an image into distinct parts
• Operation:
– setting all gray levels below a certain level to
“zero”, and above a certain level to a maximum
brightness

240-373 Image Processing 10

 Code for Thresholding

for (row=0; row<rowmax; roww++)

for (col=0; col=colmax; col++)
{
if (image[row,col] > threshold)
image[row,col] = MAX;
else
image[row,col] = MIN;
}

240-373 Image Processing 11

 Intensity Transformation
1
Adjusted Image

0.5

0
0 0.5 1

Histogram Histogram

240-373 Image Processing 12

 Thresholding Errors

• Rarely is it possible to identify a perfect gray

level break, what we want to be background
pixels become foreground or vice versa
– Type 1: not all pixels caught that should be
included
– Type 2: some pixels caught should not be
included in the group

240-373 Image Processing 13

An image before and after thresholding

240-373 Image Processing 14

 Bunching (Quantizing)

• Use:
– to reduce the number of different gray level in an
image
– to segment an image
– to remove unwanted gray level degradation
• Operation:
– Close gray levels are combined, thus removing
unwanted variations in data

240-373 Image Processing 15

 Bunching (Quantizing)

• Method 1: inspecting histogram and

combining close group into single gray level
• Method 2: identifying a set of gray levels
allowed in the final image, then changing the
gray level in every pixel to its nearest allowed
value

240-373 Image Processing 16

 Bunching Example

1 8 4 3 6 2 5 2 8 4 6 2 5 0 9 3 3 6 3 6 3 9 3 6 3 6
0 3 8 3 6 5 4 0 3 8 3 8 7 0 3 9 3 6 6 3 0 3 9 3 9 6
3 8 4 7 6 2 8 3 7 3 7 6 1 3 9 3 6 6 3 9 3 6 3 6 6 0
0 9 8 0 5 4 8 5 9 3 7 2 9 0 9 9 0 6 3 9 6 9 3 6 3 9

0 **** 0 ******
1 ** 1
2 ***** 2
3 ********* 3 *******************
4 ***** 4
5 ***** 5
6 ***** 6 ***************
7 ***** 7
8 ********* 8
9 *** 9 ************
240-373 Image Processing 17
 Bunching Code

for (row=0; row<rowmax; row++)

for (col=0; col<colmax; col++)
{
image[row,col] = bunchsize*((int)image[row,col]/bunchsize);
}

bunchsize = number of levels to be grouped into one

240-373 Image Processing 18

 Splittings

• Use:
– to increase the different two groups of gray levels
so that the contrast between segments compose
of one group of the other is enhanced
• Operation:
– rounding the gray levels up if they are in the
range and down if they are in another

240-373 Image Processing 19

 Splitting Example

• The characters on a car number-plate are at

gray level 98
• The background of the characters is at gray
level 99
• Pushing 98 down to 80 and pushing 99 up to
120 will give the picture a better contrast
around the number plate

Question: How to find a good splitting level?

240-373 Image Processing 20

 Automatic Selection of Splitting Level

• Use:
– to find the best gray level for splitting--usually for
thresholding to black and white
• Operation: g
– Let t ( g )   f (i )
i 0

where f ( g ) is the number of pixels at gray level g and

t ( g ) is the actual number of pixels at gray level g or less
in the image

240-373 Image Processing 21

Automatic Selection of Splitting Level

– Let P=NxM = the number of pixels under consideration

– Let m(g) = mean gray level for only those pixels containi
ng gray level between zero and g, i.e.
g

 g . f (i )
m( g )  i 0
t( g )

If the maximum number of gray level is G (G=0,…,G-1) then e

valuate the following equation (T = splitting threshold)

 t( g ) 2
T  max    m( g )  m(G  1)   1
 P  t( g ) 
A Processing
240-373 Image B 22
 Example
Histogram f(g) t(g) g.f(g) g.f(g) m(g) A B A*B

0 **** 4 4 0 0 0 0.08 23.04 0.18

1 ** 2 6 2 2 0.3 0.13 20.25 2.83
2 ***** 5 11 10 12 1.1 0.27 13.69 3.70
3 ********* 9 20 27 39 2 0.63 7.84 4.94
4 ***** 5 25 20 59 2.4 0.93 5.76 5.36
5 ***** 5 30 25 84 2.8 1.36 4.00 5.44
6 ***** 5 35 30 114 3.3 2.06 2.25 4.64
7 ***** 5 40 35 149 3.7 3.33 1.21 4.03
8 ********* 9 49 72 221 4.5 16.33 0.09 1.47
9 *** 3 52 27 248 4.8 -INF-
T = max(A*B) - 1 = 4
240-373 Image Processing 23