Image Processing Reference
In-Depth Information
Now we can't see
anything
! This is because there are only two brightness levels in the
image (it wasn't acquired by a camera with exponential performance). In order to show up
more
clearly
what happens to images, we need to be able to manipulate their histograms.
Intensity normalisation
stretches
a picture's histogram so that
all
available brightness
levels are used. Having
shifted
the origin to 0, by subtracting the minimum brightness, we
then
scale
up the brightness, by multiplying by some fraction of full range. It's also called
histogram normalisation
. Let's say we have 8-bit pixels, giving 256 brightness levels
(0..255), our function is:
normalise(pic):= min_val
←
Find maximum
min(pic)
max_val
←
Find minimum
..max(pic)
range
←
..max_val-min_val
Find range of intensity
for x
∈
0..cols(pic)-1
for y
∈
0..rows(pic)-1
newpic
y,x
←
floor
255
range
+0.000001
Map intensity values
(pic -min_val)
y,x
newpic
So let's normalise the eye image:
new_pic:=normalise(eye)
. This makes maximal
use of the available grey levels.
0
27
1
32
2
67
3
135
4
135
5
111
6
111
7
81
8
27
9
73
0
1
2
3
4
5
6
7
8
9
10
11
12
81
67
92
135
97
111
111
27
27
81
92
111
111
127
127
135
32
32
32
92
92
105
111
135
135
111
27
24
32
92
92
111
111
127
122
84
32
27
81
92
111
111
111
135
119
67
27
86
81
108
new_pic =
113
111
100
122
84
27
73
111 111
111
111
97
135
127
67
27
92
92 111
111
84
108
149
97
27
65
92
113 111
97
32
111
135
67
27
27
86
119
97
92
73
127
122
67
27
27
100
119
97
35
86
119
84
67
67
24
81
92
97
92
86
108
67
32
27
27
32
32
73
73
(a)
(b)
Let's see the normalised histogram:
n_hist:=histogram(new_pic)
