Gholamreza Anbarjafari, PhD

Video Lecturers on Digital Image Processing

Digital Image
Processing
Contrast Enhancement: Part II
 Gholamreza Anbarjafari, PhD
Video Lecturers on Digital Image Processing

Histogram Processing

Histogram : is the discrete function h(rk)=nk , where rk is the kth

gray level in the range of [0, L-1] and nk is the number of pixels
having gray level rk.

Normalized histogram : is p(rk)=nk/n, for k=0,1,…,L-1 and p(rk)

can be considered to give an estimate of the probability of
occurrence of ray level rk.
 Gholamreza Anbarjafari, PhD
Video Lecturers on Digital Image Processing

Histogram Equalization

Histogram equalization : is a method which increases the

dynamic range of the gray-levels in a low-contrast image to cover
full range of gray-levels.

How-to-Do: is achieved by having a transformation function

which is the Cumulative Distribution Function (CDF) of a given
PDF of gray-levels in a given image.
 Gholamreza Anbarjafari, PhD
Video Lecturers on Digital Image Processing

Histogram Equalization

Histogram equalization : the new intensity value of pixel x is

calculated by:

 cdf  x   min cdf 

I  x   round    L  1 
 1  min cdf 
 Gholamreza Anbarjafari, PhD
Video Lecturers on Digital Image Processing

Histogram Equalization

Histogram equalization : the probability function of the output

levels is uniform.

Note : the transformation function is simply the CDF.

 Gholamreza Anbarjafari, PhD
Video Lecturers on Digital Image Processing

Histogram Equalization
16
x 10
4

14

Histogram Equalization
12

10

0 50 100 150 200 250

4
x 10

16

14

12

10

0 50 100 150 200 250

 Gholamreza Anbarjafari, PhD
Video Lecturers on Digital Image Processing

Histogram Equalization

(a) A face image

from the CALTECH
face database, (b)
its histogram, (c)
the equalized face
image using HE, (d)
and its respective
histogram.
 Gholamreza Anbarjafari, PhD
Video Lecturers on Digital Image Processing

Singular Value Equalization

Singular value decomposition : any matrix, A, can be written

as multiplication of two orthogonal square matrices, U and V, and
a matrix containing the sorted singular values on its main
diagonal, Σ.

A=UΣV T
 Gholamreza Anbarjafari, PhD
Video Lecturers on Digital Image Processing

Singular Value Equalization

Note : as σ1 is much bigger than other σs then changing it will

affect on the reconstructed image, i.e. changing σ1 will directly
change the luminance of the image.
 Gholamreza Anbarjafari, PhD
Video Lecturers on Digital Image Processing

Singular Value Equalization

G(0.5, 1) : is a synthetic intensity matrix whose pixel values

have Gaussian distribution with mean of 0.5 and variance of 1
with the same size of the original image.

ξ: is ratio of the largest singular value of the generated

normalized matrix over a normalized image.



max G  0.5, 1  
max   A 
 Gholamreza Anbarjafari, PhD
Video Lecturers on Digital Image Processing

Singular Value Equalization

Equalized Im age  T
 U A  A VA 
 Gholamreza Anbarjafari, PhD
Video Lecturers on Digital Image Processing

Singular Value Equalization

Low contrast Histogram equalization Singular value

equalization
 Gholamreza Anbarjafari, PhD
Video Lecturers on Digital Image Processing

Summary
•We have looked at:
– How histogram equalization works.
– What is SVD?
– How SVE works

•Next time we will continue our talk

about image enhancement in spatial
domain