See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/350735568

A Review on Image Enhancement Techniques using Modified Approach of

Histogram Equalization

Article · April 2021

CITATIONS READS

2 136

6 authors, including:

Thamizharasan Viswanathan
Erode Sengunthar Engineering College
20 PUBLICATIONS 73 CITATIONS

SEE PROFILE

All content following this page was uploaded by Thamizharasan Viswanathan on 08 April 2021.

The user has requested enhancement of the downloaded file.

 INTERNATIONAL JOURNAL FOR RESEARCH & DEVELOPMENT IN Volume-13,Issue-5(May-20)
TECHNOLOGY ISSN (O) :- 2349-3585

A Review on Image Enhancement Techniques using

Modified Approach of Histogram Equalization
__________________________________________________________________________________________
Dr.V.Premchandran1, P.Logamurthy2, M.Karthikkumar3, V.Thamizharasan4, G.Lingeswaran 5,
M.Ramya6
1,2,3,4,5,6
Assistant Professor, Dept. of ECE,Erode Sengunthar Engineering College, Perundurai, Erode District

ABSTRACT:Image enhancement in one among in image One of the most popular contrast enhancement methods is
processing and it used to improvise image visibility. Contrast histogram equalization (HE). The mechanism of HE is to
is an important feature in an image and it should be in transform the gray levels of an image to a uniform histogram
balanced & equally distributed throughout the image. based on the probability of occurrence of gray levels in an
Contrast enhancement is one of the image enhancement input image. In general, HE will flatten out the probability
operation and it is used regularize the contrast of an image. distribution of an image and increase its dynamical range.
Histogram equalization is traditional approach in contrast However, the effectiveness of HE depends on the contrast of
enhancement technique. Even though Histogram the original image. The lower the contrast, the more the
equalization is traditional one and it has been modified by effectiveness is. Since its simplicity and ease to implement, it
many researchers in current years and produced good is usually applied in many areas including medical image
results in contrast enhancement. In this paper various type processing, texture synthesis, and speech recognition.
of modified histogram of recent years are discussed with it Histogram equalization is a well-known contrast enhancement
features. technique due to its strong performance on almost all types of
INTRODUCTION: image. Generally, histogram equalization can be categorized
Image enhancement is a process involving changing the into two main processes: global histogram equalization (GHE)
pixels’ intensity of the input image, so that the output image and local histogram equalization (LHE). In GHE, the
should subjectively looks better. The purpose of image histogram of the whole input image is used to compute a
enhancement is to improve the interpretability or perception of histogram transformation function. As result, the dynamic
information contained in the image for human viewers, or to range of the image histogram is flattened and stretched, and
provide a “better” input for other automated image processing the overall contrast is improved. The computational
systems. There are many image enhancement methods have complexity of GHE is comparatively low, making GHE an
been proposed. A very popular technique for image attractive tool in many contrast-enhancement applications. The
enhancement is histogram equalization (HE). This technique is major drawbacks of GHE are that it cannot adapt the local
commonly employed for image enhancement because of its information of the image and preserve the brightness of the
simplicity and comparatively better performance on almost all original image. In contrast, LHE uses a sliding window
types of images. The operation of HE is performed by method, in which local histograms are computed from the
remapping the gray levels of the image based on the windowed neighborhood to produce a local intensities
probability distribution of the input gray levels. It flattens and remapping for each pixel. The intensity of the pixel at the
stretches the dynamic range of the image’s histogram and center of the neighborhood is changed according to the local
resulting in overall contrast enhancement Contrast intensity remapping for that pixel. LHE is capable of
enhancement plays an important role in the field of image producing great contrast results but is sometimes thought to
processing applications. The objective of this method is to over-enhance images. It also requires more computation than
make an image clearly recognized for a specific application. other methods because a local histogram must be built and

188
All rights reserved by www.ijrdt.org
Paper Title:-A Review on Image Enhancement Techniques using Modified Approach of Histogram Equalization

processed for every image pixel. can be calculated as follow:

1)BRIGHTNESS PRESERVING DYNAMIC starti= i−1
k=1 rangek+1
HISTOGRAM EQUALIZATION (BPDHE): endi= 𝑖𝑘=1 rangek
D. Equalize each partition independently.
Brightness preserving dynamic histogram equalization
The next step in BPDHE is to equalize each partition
(BPDHE) which is proposed in this paper consists of five
independently. For sub-histogram i with the range of
steps:
[starti,endi], the equalization of this section
1. Smooth the histogram with Gaussian filter.
E. Normalize the image brightness.
2. Detection of the location of local maximums from the
In this step, the mean brightness of the input, Mi, and the
smoothed histogram.
mean brightness of the output obtained after the equalization
3. Map each partition into a new dynamic range.
process, Mo, is calculated. In order to shift back the mean
4. Equalize each partition independently.
brightness to the mean brightness of the input.
5. Normalize the image brightness.
g(x,y)=(Mi/Mo)f(x,y)
The details of each step are described in the following 2) BRIGHTNESS PRESERVING WEIGHT
subsections. CLUSTERING HISTOGRAM EQUALIZATION
A. Smooth the histogram with Gaussian filter. Here, we introduce the proposed BPWCHE method. When
Because the histogram of the digital image is normally generating the initial clusters from the original histogram, we
fluctuated and also the probability for some brightness levels assign each non-zero bin to a unique cluster. To reduce the
is missing, it is difficult to detect the local maximums of the number of clusters, we merge neighboring clusters in the
histogram without smoothing the histogram. In this step, first, original image histogram according to three criteria: cluster
the disappeared brightness level is filled up by using the linear weight, cluster weight ratio, and cluster width of two
interpolation. Then, the histogram is smoothed up by using neighboring clusters. The cluster weight criterion (W%)
one dimensional Gaussian filter. The Gaussian filter is defined utilizes the percentage of the total weight of the entire original
by the following equation: image histogram that is occupied by the current cluster. The
G(x) = exp(-x /2σ )2 2
main role of this criterion is to avoid the effects of bins of
B. Detection of the location of local maximums from the
disparate weights. When the histogram has few large bins that
smoothed histogram.
contain most of the information in the input image, previous
The smoothed histogram is only used in the process of
methods fail to enhance visualization of the image. For
splitting the original histogram. The histogram is divided into
instance, GHE, DSIHE, and BPDHE have a washed-out
sub-histograms based on local maximums. We choose to use
appearance in some images. The cluster-weight criterion
local maximums as the separating intensities rather than local
enables us to avoid this pitfall. In our algorithm, we chose a
minimums because this selection is better in maintaining the
cluster weight of W=5. Two neighboring bins the weights of
mean brightness.
which sum to less than W can be merged into one cluster. The
C.Map each partition into a new dynamic range.
cluster weight ratio criterion (R) expresses the degree of
However, the equalized version of these sub-histograms does
similarity between two neighboring clusters with respect to
not assure a very good enhancement, because sub-histograms
cluster weight. We chose R to be in the interval [R1, R2]
with small range will not be enhanced significantly by HE.
(where R1== 40:60 =0.67 and R2= 60:40 =1.5). If two
Hence, following the same concept as DHE, BPDHE spans
neighboring clusters have widely differing weights, they are
each sub-histogram first before the equalizations are taking
not merged because these clusters may contain different visual
place. Let the range of the output sub-histogram i is [starti,
information in the image. The cluster width criterion (T)
endi]. If we set the first sub-histogram of the output image is
describes the degree to which the gray levels of two
in the range of [0, range1], then the starti and endi (for i> 1)
neighboring clusters are similar. When the cluster width and

189
ISSN:-2349-3585 |www.ijrdt.org
Paper Title:-A Review on Image Enhancement Techniques using Modified Approach of Histogram Equalization

cluster weight criteria for two neighboring clusters are less Em(i) by the weights of the corresponding clusters from
than T and W, respectively, these clusters are merged into one largest to smallest, and then to the largest Em(i), for i from 1
cluster without considering the criterion R. We have chosen to D, we add one bin each. Thus, the sum of Em(i) is fixed
T=3 because we assume that three adjacent intensities will be when D is nonzero.
perceived similarly by the human eye. d)Finally, we create the transformation function.
Basic BPWCHE T(x)=starti+(endi+starti) 𝑥𝑘=𝑠𝑡𝑎𝑟𝑡 𝑛𝑘/𝑚
a)The basic steps in the new method are explained below. wherenk is number of pixels with intensity k, and M is the

Let the original image have dimensions N by M on the total number of pixels contained in that partition.

grayscale [0, L-1] and let Hn(k) denote the histogram of the 3) IMAGE CONTRAST ENHANCEMENT USING A

original image. To perform histogram weight clustering, we MODIFIED HISTOGRAM EQUALIZATION

first assume that each non-zero bin is assigned to a unique The histogram equalization is the basic technology in image

cluster. Let the i-th non-zero bin be assigned to a cluster Cn(i) processing, which is widely used to increase the integral

and its corresponding weight be Wn(i), where n is the total contrast of the image and is characterized by high efficiency.

number of non-zero bins. However, known techniques of histogram equalization have a

b) We then merge each pair of neighboring clusters when the number of disadvantages, the major one of which is a decrease

sum of their cluster weights is less than W, and when the in the contrast of small objects in the image. To address this

weight ratio, R1, R2, is in the interval [0.67, 1.5]. If the j-th disadvantage, a new technique of modified histogram

cluster satisfies above criteria.the merge will be executed as equalization is proposed based on the assessment of the two-

follows: dimensional probability distribution of brightness in the

R1=Cn(i)/Cn(i-1) or R2=Cn(i-1)/Cn(i) image. A new method of increasing the contrast of the image
c) Next, if the sum of the widths of two neighboring clusters is based on the modified equalization of histogram is also
less than T and the sum of their weights is less than W, we proposed. The proposed technique of modified histogram
then merge these clusters into one cluster without considering equalization can be recommended to increase the image
the criterion R. These clusters will be merged as in formulas contrast in automatic mode in imaging and image processing.
(2) and (3). After the above steps, we have m clusters found The methods of image enhancement by histogram
using three criteria. The main concept of the proposed transformation are widely used for images processing to
BPWCHE method is that all created clusters, m, should play increase their integral contrast and are characterized by high
equal roles, which means that the result image histogram is efficiency. The techniques of image processing by histogram
divided equally into m partitions between First bin and Last equalization provide the most effective enhancement of image
bin, which contain the lowest and highest intensities of the contrast and are widely used to improve the image quality. At
input image, respectively. Next, we will compute the number present, histogram equalization is the basic technique of image
of bins to be occupied by one cluster in the partition of the enhancement. However, known techniques of histogram
result image histogram. The essential idea is that all clusters, equalization have a number of disadvantages that substantially
m, occupy the same partition in result image histogram. limit their practical use in the automatic mode. The main
PSize= integer [(Lastbin+1-Firstbin)/m] disadvantages of the known methods of image enhancement
Em(i)=PSize by histogram equalization are an overly increase in the
D = (Lastbin+1-Firstbin) – m*PSize contrast of large-sized objects in the image. The efficiency of
whereEm(i) is the number of bins in the corresponding techniques of linear stretching (1), nonlinear stretching using
partition of the result image histogram, for the i-th cluster, and the sigmoid function (3) and gamma correction (4) for images
D is the remainder from (4). If D is nonzero, then it is enhancement depends significantly on the distribution of
necessary to alter Em(i) for each cluster. To do this, we sort image brightness and the values of parameters of the

190
ISSN:-2349-3585 |www.ijrdt.org
Paper Title:-A Review on Image Enhancement Techniques using Modified Approach of Histogram Equalization

transformation function which significantly limits the use of recommended to increase the integral contrast in imaging and
these methods. The use of modified equalization of histogram image processing in automatic mode.
based on the assessment of the two-dimensional probability 4) NEW APPROACH TO THE IMPLEMENTATION OF
distribution of brightness allows increasing the efficiency of HISTOGRAM EQUALIZATION
contrast enhancement.The proposed method for image Histogram equalization is one of the most well known
processing by modified equalization of histogram increases techniques of image processing in the spatial domain and is
the contrast of small sized objects and the integral contrast for widely applied to enhance images in various applications
all test images. At present, the solution of the task of owing to its high efficiency and simplicity. However, the
enhancement of image quality with an acceptable level of traditional definition of histogram equalization procedure does
computational costs is urgent as never before. The histogram not satisfy a number of basic requirements to image intensity
equalization is currently the basic technique of image transformations. This paper proposes a new approach to
processing. The technique of histogram equalization and its defining a histogram equalization procedure based on the
modifications are widely used for images processing to analysis of the basic requirements to the image intensity
increase image contrast and are characterized by high transformation function. Based on this approach, we propose a
efficiency. Histogram equalization techniques provide an new implementation for the histogram equalization procedure
effective increase of the generalized contrast and enhance the that satisfies the basic requirements to image intensity
objective quality of the image. However, known techniques of transformations and meets the criterion of self-duality.
histogram equalization have a number of disadvantages that 5) CONTRAST LIMITED ADAPTIVE HISTOGRAM
limit their use for image processing in the automatic mode. EQUALIZATION
Their main disadvantages are an overly increase in the contrast The CLAHE methods have produced good results on medical
of large-sized objects and also a decrease in the contrast of images. Therefore, the proposed image enhancement for 3D
small-sized objects in the image. To address these images of stereoscopic endoscopy is based on CLAHE.
disadvantages, a new technique of modified histogram There are six steps for CLAHE:
equalization was proposed. The proposed technique of Step1: The image is divided into several non-overlapping
modified histogram equalization is based on the assessment of regions of almost equal sizes. To achieve good statistical
the two dimensional probability distribution of brightness. To estimation, 64 tiles in 8 columns and 8 rows is a common
demonstrate the possibilities of the proposed technique, a new choice.
method to increasing the integral contrast of complex Step2: Histogram of each region is calculated.
monochrome images based on the modified equalization of Step3: Based on a desired limit for contrast, calculate the clip
histogram was proposed. The proposed method provides an limit. The clip limit Ncl is defined in following equation.
effective increase in the contrast of small objects in the image, Ncl = Nmc(NxNy/ Ngray)
Where Nx and Ny are the numbers of pixels in the X and Y
and also increases the integral contrast of the complex multi-
directions of the contextual region. Ngray is the number of
element image as a whole. The research of the proposed
gray level in the contextual region. Nmc is the maximum
method of modified equalization of histogram was carried out.
multiple of average pixels in each gray level of the contextual
The research was carried out by analyzing the results of the
region.
measurement of integral contrast for monochrome test images
Step4: Redistribute each histogram in such a way that its
that were processed using the known and proposed methods of
value does not go beyond the clip limit. If the number of
image enhancement. To measure the integral contrast of the
pixels is greater than Ncl, the pixels will be clipped. The extra
test images, the metrics of contrast were used. The proposed
counts, beyond the clip limit, are uniformly distributed among
technique of modified histogram equalization provides
the grayscales with counts less than the clip limit. Then, the
effective enhancement of image quality and can be

191
ISSN:-2349-3585 |www.ijrdt.org
Paper Title:-A Review on Image Enhancement Techniques using Modified Approach of Histogram Equalization

new histogram can be achieved. images without introducing unwanted artifacts, while at the
Step5: Apply histogram equalization to the gray level same time maintain the input brightness. Furthermore, similar
histogram which is limited contrast in each region. to other HE based algorithms BPDHE is easy to implement
Step6: The center points of the contextual region are regarded and can be used in real time system because of its simplicity.
as the sample points. Then, the result mapping at any pixel is BPWCHE method for contrast enhancement. BPWCHE can
interpolated from the sample mapping at the four surrounding not only preserve image brightness but can also enhance
sample points. image contrast without deterioration of visualization of the
B.Choosing reference image original image. Although previous extensions of GHE can
After processing by CLAHE, two stereo views are converted preserve image brightness, depending on the characteristics of
from RGB to YCbCrcolor space. Where Y is the intensity, and the image, they can either overdraw or fail to enhance image
Cb, Cr are the color components. Then, we calculate the contrast and visualization. BPWCHE eliminates these
average values Yavg1 and Yavg2 of Y for each image. The problems without undesirable Artifacts. BPWCHE requires no
image the average brightness value of which is major is parameters and is both simple and fully automatic, while
selected as reference image. preserving image brightness and enhancing image contrast and
C. Color matching visualization. Modified histogram equalization provides
In order to reduce the computational complexity, we correct effective enhancement of image quality and can be
colors between two stereo views based on global histogram recommended to increase the integral contrast in imaging and
matching. Histogram matching is the process of updating a image processing in automatic mode. New Approach to the
given histogram according to a prescribed one, making the Implementation of Histogram Equalization method
position and shape of both similar. As already mentioned, we implementation of the histogram equalization procedure has
have chosen one image of two stereo views as the reference low level of computation cost, effective computationally and
image Iref. In this step, the other image of two stereo views as simple to implement. The proposed definition can be
the target image Idist is transferred to the reference image Iref recommended for the implementation of histogram
by defining a mapping function M[i]. All calculation steps will equalization procedure in various image processing
be conducted independently on each color channel. There are techniques. Contrast limited adaptive histogram equalization
five steps for histogram matching. algorithm produces high quality 3D images by improving
There are five steps for histogram matching. image contrast, enhancing the details of the image obviously,
Step1: Calculate histograms Hdist[i] and Href [i] for both removing the color cast and improving consistency of two
images Idist and Iref . stereo views. However, compared with HE, AHE and CLHE,
Step2: Normalize histogram, so that individual bins will then the computational complexity of the proposed method is high.
represent the fraction of the total number of events assigned to In the future, our research will be focused on decreasing
the entire histogram. computational complexity.
Step3: Compute the cumulated histograms CD[i] and CR[i] REFERENCE:
for both images Idist and Iref 1. Sergei Yelmanov, YuriyRomanyshyn,” Image Contrast
Step4: Get the transform function M[i] according to Enhancement Using a Modified Histogram Equalization”
cumulated Histogram IEEE Second International Conference on Data Stream
Step5: The color corrected image Inew is calculated by Mining & Processing (2018), Lviv, Ukraine.
applying the mapping function M[i] to Idist 2.YuanHai, Ling Li and JiaGu, “Image Enhancement Based
DISCUSSION AND CONCLUSIONS on Contrast Limited Adaptive Histogram Equalization for
BPDHE is that there is no parameter need to be tuned. 3D Images of Stereoscopic Endoscopy”(2015), Proceeding of
Experimental results shows that BPDHE can enhance the the 2015 IEEE International Conference on Information and

192
ISSN:-2349-3585 |www.ijrdt.org
Paper Title:-A Review on Image Enhancement Techniques using Modified Approach of Histogram Equalization

Automation Lijiang, China.

3. Haidi Ibrahim, “Brightness Preserving Dynamic Histogram
Equalization for Image Contrast Enhancement”(2007), IEEE
Transactions on Consumer Electronics.
4. A. Rafael C. Gonzalez, and Richard E. Woods, “Digital
Image Processing,” 2nd edition, Prentice Hall, 2002.
5. W. K. Pratt, Digital image processing: PIKS Scientific
inside, 4th ed.,PixelSoft Inc., Los Altos, California, 2017.
6. R. C. Gonzalez and R. E. Woods, Digital Image Processing,
4nd Edition, London: Pearson, 2017.
7. Y. Rao, L. Chen, “A Survey of Video Enhancement
Techniques”,Journal of Information Hiding and Multimedia
Signal Processing, Vol. 3, No. 1, pp. 71-99, January, 2012.
8. T. Pietkiewicz, J. Matuszewski, “Recognition of Maritime
Objects on FLIR Images Using the Method of Eigenimages”,
in Proc. of the 14th Int. Conf. on Modern Problems of Radio
Eng., Telecommunications and Comp. Science (TCSET),
Lviv, Ukraine, 2018, pp. 921-929.
9. R. A. Hummel, “Histogram modification techniques”,
Computer Graphics and Image Processing, vol.4, issue 3, pp.
209–224, 1975.
10. R. Ghabousian and N. Allahbakhshi, “Survey of
ContrastEnhancement Techniques based on Histogram
Equalization,” Int. J.Rev.Life Sci., vol. 5, no. 8, pp. 901–908,
2015.
11. G. A. Shah end al., “A review on image contrast
enhancement techniques using histogram equalization”, in
International Scientific, Lahore, Pakistan, v. 27(2), pp. 1297-
1302, 2015.
12. Y.-T. Kim, “Contrast enhancement using brightness
preserving bi-histogram equalization,” IEEE Trans. on
Consumer Electronics,vol. 43, no. 1, pp. 1-8, Feb. 1997.

193
View publication stats
ISSN:-2349-3585 |www.ijrdt.org