Histogram: CII4F3 / Digital Image Processing

CII4F3 / Digital Image Processing
Histogram
Dr.Eng. Ir. Wikky Fawwaz Al Maki, S.T., M.Eng., Dr. Ema Rahmawati
Intelligent Computing and Multimedia (ICM)
1 11/02/2020 CIG4E3 / Pengolahan Citra Digital

Overview
Image Enhancement
Basic of Histogram
Histogram Calculation
Histogram Equalization

IMAGE ENHANCEMENT

Image Enhancement
“An image manipulation process to obtain an

enhanced image for a specific purposes”

Image Enhancement Approach
Frequency domain-based method

– Manipulating the frequency representation of an image
– Example: Applying FT-based transformation to an image
Spatial domain-based method

– Manipulating the image pixels directly
– Example: histogram based method

Image Histogram
Histogram : gray level value
RGB  every color channel
Equation:
nk
pr (rk )  ; 0  rk  1; k  0,1,..., L  1
n
– L: number of level
– pr(rk): probability the kth level occurs
– nk: number of occurrence of kth level
– n: total number of pixels in an image
Histogram
77 11/02/2020
11/02/2020 CIG4E3 / Pengolahan Citra Digital
BASIC OF HISTOGRAM

Histogram : Definition
Histograms plots how many times (frequency) each
intensity valuein image occurs
Example:
Image (left) has 256 distinct gray levels (8 bits)
Histogram (right) shows frequency (how many times) each gray level occurs

Many cameras display real time histogra
ms of scene
Helps avoid taking over‐exposed pictures
Also easier to detect types of processing p
reviously applied to image

E.g. K = 16, 10 pixels have
intensity value = 2
Histograms: only statistical
information
No indication of location of pixels

Different images can have same histogram
3 images below have same histogram
Half of pixels are gray, half are white
Same histogram = same statisics
Distribution of intensities could be different
Can we reconstruct image from histogram? No!


A histogram for a grayscale image with intensity values in range
would contain exactly K entries
E.g. 8‐bit grayscale image,
Each histogram entry is defined as:
E.g:
Formal definition

Interpreting Histogram
Log scale makes low values more visible
Difference between darkest and lightest

Use of Histogram
Histograms help detect image acquisition issues
Problems with image can be identified on histogram
Over and under exposure
Brightness
Contrast
Dynamic Range
Point operations can be used to alter histogram.
Addition
Multiplication
Exp and Log
Intensity Windowing (Contrast Modification)

Image Brightness
Brightness of a grayscale image is the average intensity
of all pixels in image

Detecting Bad Exposure using Histograms
Exposure? Are intensity values spread (good) out or bunched up (bad)

Image Contrast
The contrast of a grayscale image indicates how easily objects in
the image can be distinguished
High contrast image: many distinct intensity values
Low contrast: image uses few intensity values

Histogram and Image Contrast
Good Contrast?
Widely spread intensity values

+ large difference between
min and max intensity values

Contrast Equation
Many different equations for contrast exist
Examples:
Michalson’s equation for contrast

Contrast Equation : Problem
These equations work well for simple images with 2
luminances (i.e. uniform foreground and background)
Does not work well for complex scenes with many
luminances orif min and max intensities are small

Histograms and Dynamic Range
Dynamic Range: Number of distinct
pixels in image
Difficult to increase image dynamic
range (e.g. interpolation)
HDR (12‐
14 bits) capture typical, then down‐
sample

High Dynamic Range Imaging
High dynamic range means very bright and very dark parts in a
single image (many distinct values)
Dynamic range in photographed scene may exceed number of
available bits to represent pixels
Solution:
Capture multiple images at different exposures
Combine them using image processing

Detecting Image Defects using Histograms
No “best” histogram shape, depends on application
Image defects
Saturation: scene illumination values outside the sensor’s range are
set to its min or max values => results in spike at ends of histogram
Spikes and Gaps in manipulated images (not original). Why?

Image Defects: Effect of Image Compression
Histograms show impact of image compression
Example: in GIF compression, dynamic range is reduced to only few intensities
(quantization)
Original Histogram
Histogram after GIF conversion

Original Image
Fix? Scaling image by 50% and Interpolating

values recreates some lost colors
But GIF artifacts still visible

Effect of Image Compression
Example: Effect of JPEG
compression on line graphics
JPEG compression designed
for color images

Large Histograms: Binning
High resolution image can yield very large histogram
Example: 32‐bit image = 232= 4,294,967,296 columns
Such a large histogram impractical to display
Solution? Binning!
Combine ranges of intensity values into histogram columns

Calculating Bin Size

Typically use equal sized bins
Bin size?
Example: To create 256 bins from 14‐bit
image

HISTOGRAM EQUALIZATION

Histogram equalization
Employing a transformation to original image histogram to obtain a
new histogram with uniform distribution (linearization)
Basic concept: transforming the probability density function to

uniform density  continuous form
Discrete form

Equalization in continuous domain
 dr 
Histogram : p s ( s )   pr ( r ) 
 ds  r T 1 ( s )
r
Transforma si : s  T ( s )   pr ( w)dw ; 0  r 1
0
Uniform :
 1 
p s ( s )   pr ( r )   1 r T 1 ( s )  1 0  s 1
 pr (r )  r T 1 ( s )

Equalization in continuous domain
pr(r) s=T(r) ps(s)
2 1.0 1.0
0.8
0.6
1 0.5
0.4
0.2
r r s
0 0 0
1 2 0.2 0.4 0.6 0.8 1.0 0.5 1.0
Original probability density function Transformation function Resulting uniform density

Equalization in the discrete form
k nj k 0  rk  1
sk  T (rk )     pr ( r j )
j 0 n j 0 k  0,1,..., L  1
1
rk  T ( sk ) 0  sk  1

Example
Image with size 64x64 pixel (8 gray level
value) has the following distribution: Histogram:
rk nk pr(rk)=nk/n 0,3
r0=0 790 0,19 0,25
probability (p r (rk))
r1=1/7 1023 0,25 0,2
r2=2/7 850 0,21 0,15
r3=3/7 656 0,16 0,1
r4=4/7 329 0,08 0,05
r5=5/7 245 0,06 0

0 1/7 2/7 3/7 4/7 5/7 6/7 1
r6=6/7 122 0,03 gray level (rk)
r7=1 81 0,02
Transformation function
0
s0  T (r0 )   pr (rj )  pr (r0 )  0.19
j 0
1
s1  T (r1 )   pr (rj )  pr (r0 )  pr (r1 )  0.44
j 0
2
s2  T (r2 )   pr (rj )  pr (r0 )  pr (r1 )  pr (r2 )  0.65
j 0
3 4
s3  T (r3 )   pr (rj )  0.81; s4  T (r4 )   pr (r j )  0.89
j 0 j 0
5 6
s5  T (r5 )   pr (rj )  0.95; s6  T (r6 )   pr (rj )  0.98
j 0 j 0
7
s7  T (r7 )   pr (rj )  1.00
j 0

1,2
1
transformed value (s k)
0,8
0,6
0,4
0,2
0
0 1/7 2/7 3/7 4/7 5/7 6/7 1
gray level (rk)

8 gray level sk is rounded to the nearest value
s0 = 0.19  1/7
s1 = 0.44  3/7
s2 = 0.65  5/7
s3 = 0.81  6/7
s4 = 0.89  6/7
s5 = 0.95  1
s6 = 0.98  1
s7 = 1.00  1

Mapping
Only 5 gray level available at uniform histogram
– r0 (790 pixel)  s0 = 1/7
– r1 (1023 pixel)  s1 = 3/7
– r2 (850 pixel)  s2 = 5/7
– r3 (656 pixel), r4 (329 pixel)  s3 = 6/7
38 – r11/02/2020
5 (245 pixel),r6 (122 pixel),r7 (81 pixel)  s4 = 7/7
CIG4E3 / Pengolahan Citra Digital
Uniform distribution?
Histogram : approximation to probability density function
It is difficult to obtain a uniform distribution

Histogram table
rk nk pr(rk)=nk/n Sk Sk x 7 Normal(Sk)
r0=0 790 0,19 0,19 1,33  1 s0=1/7
r1=1/7 1023 0,25 0,44 3,08  3 s1=3/7
r2=2/7 850 0,21 0,65 4,55  5 s2=5/7
r3=3/7 656 0,16 0,81 5,67  6 s3=6/7
r4=4/7 329 0,08 0,89 6,23  6 s4=6/7
r5=5/7 245 0,06 0,95 6,65  7 s5=7/7
r6=6/7 122 0,03 0,98 6,86  7 s6=7/7
r7=1 81 0,02 1,00 7 s7=1

Histogram equalization result

Histogram equalization result 2

HISTOGRAM SEPCIFICATION

Histogram Specification
Drawback of histogram equalization : the histogram can’t be
obtained based on purposes
In fact, sometimes we need to stress a specific interval of image

gray levels  histogram specification

Histogram Specification
A 64x64 pixel image with 8 gray level
has the following distribution : Image histogram:
rk nk pr(rk)=nk/n
0,3
r0=0 790 0,19
0,25
r1=1/7 1023 0,25
probability (p r (rk))
0,2
r2=2/7 850 0,21
0,15
r3=3/7 656 0,16
0,1
r4=4/7 329 0,08
0,05
r5=5/7 245 0,06
0
r6=6/7 122 0,03 0 1/7 2/7 3/7 4/7 5/7 6/7 1
r7=1 81 0,02 gray level (rk)

The desired histogram
zk pz(zk)
0,35
z0=0 0,00
0,30
z1=1/7 0,00
probability (p z (z k))
0,25
z2=2/7 0,00
0,20
z3=3/7 0,15 0,15
z4=4/7 0,20 0,10
z5=5/7 0,30 0,05
z6=6/7 0,20 0,00

0 1/7 2/7 3/7 4/7 5/7 6/7 1
z7=1 0,15
gray level (z k)

Step 1 : Histogram equalization
After histogram equalization, we get :
0,3
rjsk nk ps(sk)
0,25
r0s0=1/7 790 0,19
probability (p s(sk))
0,2
r1s1=3/7 1023 0,25
0,15
r2s2=5/7 850 0,21
0,1
r3,r4 s3=6/7 985 0,24
0,05
r5,r6,r7 s4=7/7 448 0,11
0
0 1/7 2/7 3/7 4/7 5/7 6/7 1
gray level (sk)

Step 2: Find the transformation function
k
vk  G ( z k )   p z ( z j )
j 0
v0 = G(z0) = 0,00
v1 = G(z1) = 0,00
v2 = G(z2) = 0,00
v3 = G(z3) = 0,15
v4 = G(z4) = 0,35
v5 = G(z5) = 0,65
v6 = G(z6) = 0,85
v7 = G(z7) = 1,00
rk nk pr(rk)=nk/n Sk Sk x 7 Normal(Sk)
r0=0 790 0,19 0,19 1,33  1 s0=1/7
r1=1/7 1023 0,25 0,44 3,08  3 s1=3/7
r2=2/7 850 0,21 0,65 4,55  5 s2=5/7
r3=3/7 656 0,16 0,81 5,67  6 s3=6/7
r4=4/7 329 0,08 0,89 6,23  6 s4=6/7
r5=5/7 245 0,06 0,95 6,65  7 s5=7/7
r6=6/7 122 0,03 0,98 6,86  7 s6=7/7
r7=1 81 0,02 1,00 7 s7=1
Perform histogram equalization to the desired histogram
zk pz(zk) Vk Vk x 7 Normal(Vk)
z0=0 0,00 0,00 0,00 v0=0
z1=1/7 0,00 0,00 0,00 v1=0
z2=2/7 0,00 0,00 0,00 v2=0
z3=3/7 0,15 0,15 1,05  1 v3=1/7
z4=4/7 0,20 0,35 2,45  2 v4=2/7
z5=5/7 0,30 0,65 4,45  4 v5=4/7
z6=6/7 0,20 0,85 5.95  6 v6=6/7
z7=1 0,15 1,00 7 v7=1
Graph of Transformation function
1,2
1
transformation (v k)
0,8
0,6
0,4
0,2
0
0 1/7 2/7 3/7 4/7 5/7 6/7 1
gray level (z k)

Step 3: Apply inverse of G to the level of histogram
equalization
Mapping the value of sk to the nearest values of G(zk)
s0 = 1/7 ≈ 0.14  G(z3) = 0.15; z3 = 3/7

s1 = 3/7 ≈ 0.43  G(z4) = 0.35; z4 = 4/7
s2 = 5/7 ≈ 0.71  G(z5) = 0.65; z5 = 5/7
s3 = 6/7 ≈ 0.86  G(z6) = 0.85; z6 = 6/7
s4 = 1  G(z7) = 1.00; z7 = 1

Step 4: Map the value of rk to zk
Notice the mapping from original histogram to histogram equalization
r0 = 0  z3 = 3/7
r1 = 1/7  z4 = 4/7
r2 = 2/7  z5 = 5/7
r3 = 3/7  z6 = 6/7
r4 = 4/7  z6 = 6/7
r5 = 5/7  z7 = 1
r6 = 6/7  z7 = 1
r7 = 1  z7 = 1

Result
zk nk pz(zk)=nk/n 0,30
r0=0 0 0 0,25
probability (p z (z k))
r1=1/7 0 0 0,20
r2=2/7 0 0 0,15
0,10
r3=3/7 790 0,19
0,05
r4=4/7 1023 0,25
0,00
r5=5/7 850 0,21
0 1/7 2/7 3/7 4/7 5/7 6/7 1
r6=6/7 985 0,24 gray level (z k)
54
r7=1
11/02/2020
448 0,11 CIG4E3 / Pengolahan Citra Digital
Algorithm: 512 x 512 Image
Var x,y,i,minval,minj,j : integer; Histspec : array[0..255] of integer;
Invhist : array[0..255] of integer; Sum : real;
Begin
Hist_Equalization(Image) {equalisasi histogram}
For i:= 0 to 255 do {histogram yang dispesifikasikan telah disimpan di spec}
Sum:= 0.0;
For j:= 0 to i do Sum := sum + spec[j]
Histspec[i] = round(255 * sum)
Endfor {didapat fungsi transformasi}
for i:= 0 to 255 do {pemetaan histogram}
minval := abs(i – histspec[0]; minj := 0;
for j:= 0 to 255 do
if abs(i – histspec[j]) < minval then
minval := abs(i – histspec[j])
minj := j
endif
invhist[i]:= minj
endfor
endfor
for y:= 0 to 511 do {ubah nilai tiap pixel pada citra}
for x:= 0 to 511 do image[x,y] = invhist[image(x,y)]

Example 1 : histogram specification

Example 2 : histogram specification

Example 3 : Histogram Specification

Method for Histogram specification
559 11/02/2020
Local enhancement
Histogram equalization and specification : global (applied to all of
the pixels)
Sometimes we need an enhancement in specific area of the image

THANK YOU

Histogram: CII4F3 / Digital Image Processing

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Histogram: CII4F3 / Digital Image Processing

Uploaded by

Copyright:

Available Formats

CII4F3 / Digital Image Processing

Intelligent Computing and Multimedia (ICM)

1 11/02/2020 CIG4E3 / Pengolahan Citra Digital

2 11/02/2020 CIG4E3 / Pengolahan Citra Digital

3 11/02/2020 CIG4E3 / Pengolahan Citra Digital

“An image manipulation process to obtain an

4 11/02/2020 CIG4E3 / Pengolahan Citra Digital

Frequency domain-based method

Spatial domain-based method

5 11/02/2020 CIG4E3 / Pengolahan Citra Digital

8 11/02/2020 CIG4E3 / Pengolahan Citra Digital

9 11/02/2020 CIG4E3 / Pengolahan Citra Digital

10 11/02/2020 CIG4E3 / Pengolahan Citra Digital

11 11/02/2020 CIG4E3 / Pengolahan Citra Digital

12 11/02/2020 CIG4E3 / Pengolahan Citra Digital

13 11/02/2020 CIG4E3 / Pengolahan Citra Digital

Difference between darkest and lightest

14 11/02/2020 CIG4E3 / Pengolahan Citra Digital

15 11/02/2020 CIG4E3 / Pengolahan Citra Digital

16 11/02/2020 CIG4E3 / Pengolahan Citra Digital

17 11/02/2020 CIG4E3 / Pengolahan Citra Digital

18 11/02/2020 CIG4E3 / Pengolahan Citra Digital

Widely spread intensity values

19 11/02/2020 CIG4E3 / Pengolahan Citra Digital

20 11/02/2020 CIG4E3 / Pengolahan Citra Digital

21 11/02/2020 CIG4E3 / Pengolahan Citra Digital

22 11/02/2020 CIG4E3 / Pengolahan Citra Digital

23 11/02/2020 CIG4E3 / Pengolahan Citra Digital

24 11/02/2020 CIG4E3 / Pengolahan Citra Digital

Histogram after GIF conversion

Fix? Scaling image by 50% and Interpolating

25 11/02/2020 CIG4E3 / Pengolahan Citra Digital

26 11/02/2020 CIG4E3 / Pengolahan Citra Digital

27 11/02/2020 CIG4E3 / Pengolahan Citra Digital

28 11/02/2020 CIG4E3 / Pengolahan Citra Digital

29 11/02/2020 CIG4E3 / Pengolahan Citra Digital

Basic concept: transforming the probability density function to

30 11/02/2020 CIG4E3 / Pengolahan Citra Digital

31 11/02/2020 CIG4E3 / Pengolahan Citra Digital

pr(r) s=T(r) ps(s)

Original probability density function Transformation function Resulting uniform density

32 11/02/2020 CIG4E3 / Pengolahan Citra Digital

33 11/02/2020 CIG4E3 / Pengolahan Citra Digital

r0=0 790 0,19 0,25

r2=2/7 850 0,21 0,15

r3=3/7 656 0,16 0,1

r4=4/7 329 0,08 0,05

r5=5/7 245 0,06 0

35 11/02/2020 CIG4E3 / Pengolahan Citra Digital

36 11/02/2020 CIG4E3 / Pengolahan Citra Digital

37 11/02/2020 CIG4E3 / Pengolahan Citra Digital

– r0 (790 pixel)  s0 = 1/7

– r1 (1023 pixel)  s1 = 3/7

– r2 (850 pixel)  s2 = 5/7

– r3 (656 pixel), r4 (329 pixel)  s3 = 6/7

Histogram : approximation to probability density function

It is difficult to obtain a uniform distribution

39 11/02/2020 CIG4E3 / Pengolahan Citra Digital

40 11/02/2020 CIG4E3 / Pengolahan Citra Digital

41 11/02/2020 CIG4E3 / Pengolahan Citra Digital

11/02/2020 CIG4E3 / Pengolahan Citra Digital

43 11/02/2020 CIG4E3 / Pengolahan Citra Digital

In fact, sometimes we need to stress a specific interval of image