IRJET- Histogram Specification: A Review

International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395-0056
Volume: 06 Issue: 03 | Mar 2019 www.irjet.net p-ISSN: 2395-0072
© 2019, IRJET | Impact Factor value: 7.211 | ISO 9001:2008 Certified Journal | Page 594
Histogram specification: A Review
PriyankaTyagi1, Dr. Tarun Gupta2, Abhishek Singh3
1M.Tech, Computer Science & Engineering, A.K.T.U Lucknow R.G.G.I Meerut, U.P, India
2Associate Professor, Computer Science & Engineering, R.G.G.I Meerut, U.P, India
3Assistant Professor, Electronics & Communication Engineering, J.P.I.E.T, Meerut, U.P, India
------------------------------------------------------------------------***-----------------------------------------------------------------------
ABSTRACT - Image preprocessing, Gaussian smoothing, Contrast stretching is concerned with producing and re-arrange an actual
array of pixels for object representation to enhance the noisy and low contrast images. In the suggested method after noise
smoothing & contrast stretching, image is equalized for better contrast using Histogram Equalization (HE), however, it tends to
change the mean brightness of the image, and hence not suitable for consumer electronics products, where preserving the original
brightness is essential to avoid annoying artifacts and unnatural enhancement. So this paper reviewed a novel extension of
histogram specification to overcome such drawback as HE, named (BBHE) Bi-Histogram Equalization,(DSIHE) Dualistic sub-
image Histogram Equalization and (BPHEME) Brightness Preserving Histogram Equalization with Maximum Entropy,
Experimental results show that BPHEME can not only enhance the image effectively, but also preserve the original brightness in
the form of mean and provide maximum entropy.
Key words: HE, BBHE, DSIHE, BPHEME.
1. INTRODUCTION
One of the most widely used technique for image enhancement is Histogram equalization (HE).The main disadvantage of
Histogram equalization (HE) is that it tends to change the mean brightness of the input image. In theory, the mean brightness
of its output image is always the middle gray level regardless of the input mean, because the desired histogram is flat. This is
not a desirable property in some applications where brightness preservation is necessary. Brightness preserving Bi-Histogram
Equalization (BBHE) has been proposed to overcome that problem [3]. BBHE first separates the input image’s histogram into
two by its mean, and thus two non-overlapped ranges of the histogram are obtained. Next, it equalizes the two sub-histograms
independently. It has been analyzed that BBHE can preserve the original brightness to a certain extent when the input
histogram has a quasi-symmetrical distribution around its mean [7]. Later, equal area Dualistic Sub-Image Histogram
Equalization (DSIHE) has been proposed, it claims that if the separating level of histogram is the median of the input image’s
brightness, it will yield the maximum entropy after two independent sub-equalizations [8]. DSIHE will change the brightness
to the middle level between the median level and the middle one of the input image. Nevertheless, neither BBHE nor DSIHE
could preserve the mean brightness. In the consumer electronics such as TV, medical imaging and Astronomical imaging the
preservation of brightness is highly demanded. The aforementioned algorithms (HE, BBHE, DSIHE) preserve the brightness to
some extent; however, they do not meet that desirable property quite well. In this paper, a novel enhancement method is
proposed which can yield the optimal equalization in the sense of entropy maximization, under the constraint of the mean
brightness, called Brightness Preserving Histogram Equalization with Maximum Entropy (BPHEME). BPHEME, together with
the aforementioned algorithms, is essentially a kind of histogram specification [1] in general, except that different “ideal”
histograms are employed in different algorithms. In the next section, histogram specification will be reviewed, and HE, BBHE,
DSIHE, will be quickly introduced as special cases of histogram specification. The proposed algorithm, BPHEME, will be
presented, which is “ideal” in the sense of maximum entropy with an invariant mean brightness. Some noise occurs during
image acquired by modern sensors consist of thermal noise, amplifier noise, photon noise, quantization noise, distortions or
shading. A linear filter is used for slowly varying noise; and. a nonlinear filter (Gaussian filter) is used for rapidly varying noise.
Histogram equalization (HE) flattens and stretches the dynamic range of the image’s histogram and results in overall contrast
improvement. However, it may significantly change the brightness of an input image and cause problem in some applications
where brightness preservation is necessary. In this paper, a novel enhancement method is proposed which can yield the
optimal equalization in the sense of entropy maximization and contrast enhancement by using Brightness Preserving
Histogram Equalization with Maximum Entropy (BPHEME) and Piecewise linear transform (PLT) respectively.

Why do we use histogram of an image?
Histogram of image gives information about image contrast.
 Dark image, Histogram will be clustered towards the lower gray level.
 Lighter image, Histogram will be clustered towards the higher gray level.
 Low contrast image, Histogram will not be spread equally.
 High contrast image, Histogram will be spread equally.
Fig.1Shows nature of Histograms
2. Histogram equalization:
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].We need to design a gray value transformation s = T(r),
based on the histogram of the input image, which will enhance the image. Histogram equalization yields an image whose pixels
are (in theory) uniformly distributed among all gray levels.
(i) T(r) is a monotonically increasing function for 0 < r < 1 (preserves order from black to white).
(ii) T(r) maps [0,1] into [0,1] (preserves the range of allowed Gray values). So the grey levels for continuous variables can be
characterized by their probability density functions p(r) and p(s). The probability density of the transformed grey level is P(s)
=P(r) (1)
To find a transformation, Let us consider the Cumulative Density Function (CDF), which is obtained by simply adding up all the
Probability density functions (PDF) S=T(r)= 
r
drrP
0
;)( ; 1 ≥ r ≥ 0 ; (2)
P(s) = [1] = 1 ; 1 ≥ s ≥0;

Fig.2 Histogram equalization for High contrast image
Drawback of Histogram Equalization:
 Change the mean brightness of the image.
 Natural look of image may be lost.
 Original image color change’s from white to black.
Fig.3 Snake color changes from white to black after histogram equalization.
3. Histogram specification (Histogram Matching):
Histogram equalization yields an image whose pixels are (in theory) uniformly distributed among all gray levels. Sometimes,
this may not be desirable. Instead, we may want a transformation that yields an output image with a pre-specified histogram.
This technique is called histogram specification. Given information are, Input image from which we can compute its
histogram and desired histogram. With the help given information derive a point operation, that maps the input image into an
output image that has the user-specified histogram.

4. Bi-Histogram Equalization (BBHE):
Let Xm is the mean of image X and the input image is decomposed into two sub-images XL and XU by its mean. Where, X=(XL U
XU) XL={X(i , j) < Xm}; XU={X(I , j)>Xm};while XL is composed of {X0 , X1 ,….Xm} and XU is composed of {Xm+1 , Xm+2 ,….XL-1} hence
the Transform functions are fL(X)=X0+(Xm-X0) CL(x) ; (3)
fU(x)=Xm+1+(XL-1-Xm-1) CU(x); (4)
Where CL(x) and CU(x) are cumulative density functions for XL and XU respectively.
Fig.4 Mean-based separated images and their equalize histogram
5. DSIHE (Dualistic sub-image histogram equalization):
In this method separating level of histogram is the median of the input image. Which provide maximum entropy after two
independent sub-equalization? Both BBHE and DSIHE is similar except that DSIHE choose to separate the histogram based on
gray level with cumulative probability density equal to 0.5 instead of the mean as in BBHE.
Fig.5 Median-based separated images and their equalize histogram.

6. Entropy (Information)
The differential entropy h(X) of a continuous random variable X with a density f(x) is defined as

s
dxxfxfxh )(log)()( (5)
Here S is the support set of the random variable. h(x) is also named as the continuous variable’s entropy [4].
7. Gaussian smoothening Filter:
Gaussian filters are a class of linear smoothing filter with the weight chosen according to the shape of a Gaussian function. The
Gaussian kernel is widely used for smoothing purpose. The Gaussian filter in the continuous space is
h(m , n)=[
√
] * [
√
] (6)
This shows that Gaussian filter is separable, rotationally symmetric in two dimensions and Fourier transform of a Gaussian
function is itself a Gaussian function. Fourier transform of a Gaussian has a single lobe in the frequency spectrum. Images are
corrupted by high –frequency noise, and the desirable feature of the image will be distributed both in the low-and-high
frequency spectrum. The single lobe in the F.T of a Gaussian means that the smoothened image will not be corrupted by
contributions from unwanted high-frequency signals, while most of the desirable signal properties will be retained. The
degree of smoothing is governed by variance (σ). A larger variance implies a wider Gaussian filter and greater smoothening.
Here a spatial image processing is used to blur image and to remove noise from a image. Blurring is used in the pre-processing
stage.
Fig.6 Gaussian smoothing image and its Histogram.
8. BPHEME: Brightness Preserving Histogram Equalization with Maximum Entropy based Histogram specification:
HE is to make the output histogram as flat as possible. A superficial reason for HE relies on that a flat histogram makes all the
gray levels uniform, and thus will cause a more comfortable perception. A further comprehension is that a uniform
distribution limited to a given range gives the maximum information, measured by entropy, so we would like to find an “ideal”
histogram (PDF) as a target one to perform the histogram specification. That ideal histogram preserves the mean brightness of
the input image, and has the maximum entropy [4]. Thus an optimal brightness preserving enhancement method using
histogram transformation may be to maximize the target histogram’s entropy under the constraints of brightness.
Mathematically speaking, we want to maximize h(f) over all probability densities f subject to some constraints:


s
dssfsff })(log)({max such that












s
s
rdsssf
sf
sf
)(
;1)(
;0)(
Now using Lagrange multiplier expression
J(f) = 
s
dssfsf ])(log)([ + λ1  
1
0
]1)([ dssf +λ2[  
1
0
)( rdsssf  ] (7)
Now ;0
)(
)(



sf
fj
f(s)= e ( λ1-1).e(s*λ2) (8)
Now from constraint (2) and eq.(8) ;
f(s) = ) )
)
(9)
µr =
) )
)
) – )
(10)
Fig.7 Plot between mean (µr) and lemda2 ( )
)= [ ) )
)
]
[
) )
)
) =1 ; (11)
( ) )
)
( ) ) ) )
= 0.5 ;
Hence f(s) ={
) )
)
) ) (12)
Thus we have cumulative histogram, or cumulative distribution ,
) ∫ ) )
) ) (13)
-100 -80 -60 -40 -20 0 20 40 60 80 100
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
lemda2
meanµr

)
)
) )
(14)
Thus we have the cumulative histogram, or cumulative distribution function, c(s) as following.
C(s) = { )
) )
) )
(15)
Fig.8 Image (DS03987) by using HE, BPHEME, BBHE , DSIHE method.
Fig.9 Histogram of image (DS03987) by using HE, BPHEME, BBHE, DSIHE method

9. Mean (Image Brightness):
The mean of a discrete random variable x is given by taking the product of each possible value of x and its probability P(x), and
then adding all these products together, giving  )(* xPx
10. Entropy (Information):


256
0
))((2log*)(
i
iPiPH . Where p(i) is the probability of occurrence of gray level (i).
11. Peak-signal-to-noise-ratio (PSNR):
PSNR is the evaluation standard of the reconstructed image quality, and is important measurement feature. PSNR is measured
in decibels (dB) and is given by: PSNR= 10 log ( 2552 /MSE).where the value 255 is maximum possible value that can be
attained by the image signal. Mean square error (MSE) is defined as Where (m*n) is the size of the original image. Higher the
PSNR value is, better the reconstructed image is.
12. Visual Quality:
By looking at the enhanced image, one can easily determine the difference between the input image and the enhanced image
and hence, performance of the enhancement technique is evaluated.
Results of test image:
Table.1 Mean
Images
ORIGINAL
(MEAN)
HE
(MEAN)
BBHE
(MEAN)
DSIHE
(MEAN)
BPHEME
( MEAN)
DS03987 104.0286 127.5565 126.9356 126.1309 103.2234
Table.2 Entropy (bits/pixel)
IMAGES ORIGINAL
IMAGE
(ENTROPY)
HE (ENTROPY) BBHE
(ENTROPY)
DSIHE
(ENTROPY)
BPHEME
(ENTROPY)
DS03987 7.5872 5.9832 7.0231 7.0238 7.4172
Table.3 Peak signal to noise ratio
IMAGE HE/PSNR(db) BBHE/PSNR(db) DSIHE/PSNR(db) BPHEME/PSNR(db)
DS03987 17.7970 19.5378 19.8126 20.9872

CONCLUSION:
BPHEME technique was proposed that specifies a histogram automatically. The method runs in 2.796 by using a 1.19 GHz PC
computer and produce satisfactorily enhance images.
REFERENCES
[1] R. C. Gonzalez and R. E. Woods, Digital Image Processing, Prentice Hall, second edition, 2001.
[2] M. Stamm, K.J.R, Liu, “Blind forensics of contrast enhancement in digital images,” IEEE International Conference on Image
Processing, San Diego, CA, Oct. 2008.
[3] C. Wang and Z. Ye, “Brightness preserving histogram equalization with maximum entropy: a variational perspective,” IEEE
Transactions on Consumer Electronics, Vol. 51, No. 4, pp. 1326-1334, November 2005.
[4] C. M. Tsai, and Z. M. Yeh, “Contrast enhancement by automatic and parameter-free piecewise linear transformation for
color images,” IEEE Transactions on Consumer Electronics, Vol. 54, No. 2, May 2008.
[5] G. J. Erickson and C. R. Smith, Maximum entropy and Bayesian methods, Kluwer Pulishers, Seattle, 1991.
[6] J.B.Zimmerman, S.M.Pizer, E.V.Staab, J.R.Perry, W.McCartney and B.C.Brenton, ‘An evaluation of the effectiveness of
adaptive histogram equalization for contrast enhancement’, IEEE Transactions on Medical Imaging, Vol.7, No.4, pp:304-312,
1988.
[7] Y.-T Kim, ‘Contrast Enhancement Using Brightness Preserving Bi-Histogram Equalization’, IEEE Transactions on Consumer
Electronics, Vol.43, No.1, pp:1-8, 1997.
[8] Y.Wang, Q.Chen and B.M.Zhang, ‘Image Enhancement based on Equal Area Dualistic Sub-image Histogram Equalization
Method’, IEEE Transactions on Consumer Electronics, Vol.45, No.1, pp:68-75, 1999.

IRJET- Histogram Specification: A Review

More Related Content

What's hot (18)

Similar to IRJET- Histogram Specification: A Review (20)

More from IRJET Journal (20)

Recently uploaded (20)

IRJET- Histogram Specification: A Review