International Journal of Scientific Engineering and Technology

Volume No.5 Issue No.6, pp: 507-511

ISSN:2277-1581
01 Nov. 2016

Novel Image Enhancement Technique using CLAHE and Wavelet Transforms

T.V. Hyma Laksmi1, T.Madhu2, K.Ch. Sri Kavya 3, Shaik. Esub Basha4
1
Assistant professor, SRKR Engineering college, Bhimavaram, Andhra Pradesh, India
2
Principal, S.I.E.T, Narsapur, Andhra Pradesh, India
3
Professor, KL university, Andhra Pradesh, India
4
PG student, SRKR Engineering college, Bhimavaram, Andhra Pradesh, India
[email protected], [email protected]
AbstractThis paper presents a new method for image
enhancement which includes both resolution enhancement
and contrast enhancement. The proposed method merges
SWT-DWT which is for resolution enhancement and CLAHESWT which is for Contrast enhancement. SWT is used in
combination with DWT for enhancing the resolution of an
image. (CLAHE) is a powerful method for contrast
enhancement.SWT is used in combination with CLAHE to
mitigate noise effects. The proposed method gives better results
than existing techniques and it is proved with PSNR, Standard
Deviation and RMSE and visual results.
KeywordsContrast
limited
adaptive
histogram
equalization(CLAHE),Discrete wavelet transform(DWT),
Stationary wavelet transform,(SWT), Bi-cubic interpolation,
Peak signal to noise ratio(PSNR)
I. Introduction
Image enhancement means to polish up the image so that
enhanced image looks better than original. Resolution
enhancement and contrast enhancement both are important
features in image processing field. Wavelet transform has
become most salient tool for image enhancement applications.
Resolution is the ability of image to show its details. Hassan
Demirel and Gholamreza Anbarjafari proposed the method[1] of
image resolution enhancement using DWT . This method
decomposes the image using DWT, and high frequency
components are bi-cubic interpolated. However in this method
noise has its significant effect. To mitigate noise increment they
introduce SWT as an intermediate stage and proposed the
method [2] of image resolution enhancement using DWT and
SWT.
Contrast enhancement is an improvement of image quality to
better and more understandable level for feature extraction or
image interpretation. Contrast enhancement techniques have
been classified into two principle groups-spatial-domain,
transform-domain based methods. Spatial domain techniques
like logarithmic transforms, power law transform, histogram
equalization methods are based on direct manipulation of pixels
in the image plane, while transform domain techniques are based
IJSET@2016

on the manipulation of the transform of the image rather than the

image
itself.
Histogram
equalization is
a
method
of contrast adjustment [3] using the image histogram in image
processing. This method increases the global contrast of images
by applying a gray level transform which tries to flatten the
resulting histogram. Histogram equalization works best on over
or under exposed image, which has narrow contrast range. Since
the HE is applied on the whole image, the local details are not
enhanced effectively. To overcome these drawbacks, local
histogram equalization (LHE) based methods are proposed.
Pizer[4] proposed the contrast adaptive histogram equalization
(CLAHE) which is an LHE-based image enhancement method.
Histograms that are above the clip limit are clipped and
distribute to the other histograms of different regions which have
histogram height below the clip limit. In transform-domain
based methods decomposition is applied i.e. transformed into
frequency domain before image enhancement, which prevents
image artifacts and improves image quality. Huang lidong
proposed [5] CLAHE-DWT method which is one of the contrast
enhancement method and uses the combination of both DWT
and CLAHE. Drawbacks of CLAHE are overcome by this
CLAHE-DWT method.
In this paper, we introduce a new contrast and resolution
enhancement technique which can improve resolution, contrast
of image and avoid over enhancement, reduce the noise effects.
This new method combines CLAHE with SWT for Contrast
enhancement and DWT with SWT for resolution enhancement.
In this proposed method input image is given to both DWT and
SWT for resolution increment. Bi-cubic interpolation is used for
sharper details. Resolution enhanced image is given to SWT for
different frequency sub-bands. SWT minimize the loss in terms
of down sampling which occur in DWT because of down
sampling. CLAHE is applied to low frequency components only.
This mitigates the effect of noise during process of contrast
enhancement. Finally inverse DWT gives contrast enhanced
image. Noise effect can be further decreased by taking the
average of original image and output of inverse wavelet
transform.

doi : 10.17950/ijset/v5s11/1103

Page 251

International Journal of Scientific Engineering and Technology

Volume No.5 Issue No.6, pp: 507-511

ISSN:2277-1581
01 Nov. 2016

II . Related work

LOW RESOLUTION AND LOW

CONTRAST IMAGE

In this section we provide the work which is related to proposed

method that is about CLAHE and CLAHE-DWT.

Interpolation by 2

SWT

CLAHE:. The main steps of CLAHE are given as follows

DWT

1. Divide the input image into tiles.

2. Set the clip limit, clip the histograms above clip limit and
redistribute to other sub-blocks.

LL

LH

LOW RESOLUTION
LOW
LOW
RESOLUTION
RESOLUTION
AND
AND
AND LOW CONTRAST
LL1
LOW
LOW
CONTRAST
CONTRAST
IMAGE
IMAGE
LH1
HH
HL
IMAGE

3. Apply histogram equalization to the each tile.

4. Interpolate the neighboring tiles.

Interp
olatio
n by
2

SWTSWT
SWT

Selection of tile size and clip limit is crucial for CLAHE because
these parameters mainly control image quality. These parameters
of CLAHE are determined based on image entropy . Image
entropy is proportional to the distribution of histogram. Selection
of tile size and clip limit follows [7] the below steps. 1. Set tile
size as [8 8], vary the clip limit and apply CLAHE to input
image.
2. Find entropy of CLAHE output, maximum entropy gives the
best clip limit.
3. Update the above obtained clip limit, now vary the tile size
from [2 2] to [32 32].
4. Repeat the step 2, maximum entropy gives the best tile size.
5. Update the tile size with value obtained above and performs
the CLAHE with these parameters.
Discrete entropy is defined as [6] follows

L
L

DWT
DWT
DWT

L L
H HL LL
LH LH HL
LH HL H
LH
LH
LH
HH
L L
L H HL LL

New
LH1

IJSET@2016

Interp
olatio
n by
2

HH
HL
LL

New
HL1

Interp
olatio
n by
2

H
H
HH
H
H

New
HH1

IDWT

HIGH RESOLUTION IMAGE

LHLHLH

---------(1)
CLAHE-DWT: This method uses the combination of CLAHE
and DWT to eliminate the drawbacks of CLAHE. Applying
CLAHE to low-frequency components only reduces the noise.
The Weighted average of the original image and reconstructed
image alleviates the contrast overstretching. Pixels with higher
intensities are enhanced less because of proportionality of
weighting factor with pixel intensities. Choosing of weighting
factor has significance and it is selected based on local entropy
increment (LEI).
III.PROPOSED METHOD:
DWT-SWT&CLAHE-SWT: This proposed method uses the
combination of DWT-SWT for resolution enhancement and
CLAHE-SWT for contrast enhancement. First, input image of
size [256256] is decomposed using both DWT and SWT using
Haar mother wavelet to get the low-frequency components and
high-frequency components individually. The Haar wavelet is
the simplest wavelet of all other wavelet families in terms of
computation of wavelet coefficients. Fig. 1 shows the flow chart
of proposed method. LL, LH, HL and HH are sub-band images
obtained by applying DWT.

HH1

HL1

LH
LH
LH

SWT

IDWT
IDWT
IDWT
LF

HF

HIGH
HIGH
HIGH
RESOLUTION
RESOLUTION
RESOLUTION
IMAGE
IMAGE
IMAGE
CLAHE

SWT
SWT
SWT
IDWT

L L L
HFHFHF
F
AVERAGE

F F
Weighted

CLAH
CLA
CLA
HE HE AND
HIGHECONTRAST
HIGH RESOLUTION
IMAGE

IDWT
IDWT
IDWT

FIG.1 Flow chart of DWT-SWT&CLAHE-SWT (PROPOSED METHOD)

doi : 10.17950/ijset/v5s11/1103

TAKE
TAKE
TAKE
AVERAGE
AVERAGE
AVERAGE
HIGH
HIGH
HIGH
CONTRAST
CONTRAST
CONTRAST AND
ANDAND
HIGH
HIGH
HIGH
RESOLUTION
RESOLUTION
RESOLUTION

Page 252

HH
HH
HH

International Journal of Scientific Engineering and Technology

Volume No.5 Issue No.6, pp: 507-511
LL is the low-frequency information obtained by applying low
pass filters and it is the approximation of original image with
half size of the original image. LH, HL and HH are the
vertical ,horizontal and diagonal information respectively. LL1,
LH1, HL1 and HH1 are the sub-band images obtained after
applying stationary wavelet transform(SWT).The stationary
wavelet transform (SWT) is used to reduce the pixel loss which
occurs in discrete wavelet transform(DWT) because of downsampling. High frequency sub-bands of DWT are interpolated
with factor 2 with the help of bi-cubic interpolation. Bi-cubic
interpolation method uses the weighted average of neighboring
sixteen pixels to estimate the intensity value of the middle pixel.
This method gives excellent results both in calculation speed and
quality of transformed image. Interpolation increases the number
of pixels in a given image. High frequency sub-bands of SWT
and interpolated sub-bands of DWT are combined. By taking the
Inverse DWT of low frequency sub-band and new sub-bands we
get the high resolution image with size [512512].
Now the image is resolution enhanced but contrast of image
is still poor .To improve the contrast of image we are using the
CLAHE. First Resolution enhanced image is given to SWT.SWT
is preferred over DWT because of its self-property of no down
sampling. This is the main advantage of SWT over DWT in
terms of pixel loss. Now, CLAHE is applied to low frequency
sub-band. This is because high frequency sub-bands contains
most of noise and by applying CLAHE to these sub-bands also
causes noise increment. Because of this CLAHE is applied to
low frequency sub-bands only. By taking the Inverse transform
of CLAHE applied sub-band and high frequency bands we get
high contrast image with size [512512].On taking the average
of low contrast image and resultant of Inverse transform noise
can be further be reduced. Finally resultant image is both
resolution enhanced and contrast enhanced.
IV. RESULTS & ANALYSIS:
In this experiment, proposed algorithm has been used for
analysis of grey and color images. . The performance of
proposed method for gray and color images is compared with
both contrast enhancement (HE, CLAHE and CLAHE-DWT)
and resolution enhancement (DWT and DWT &SWT) methods
with different parameters (peak signal to noise ratio (PSNR),
noise estimation (NE) and root mean square error (RMSE)).
Both visual results and quantitative results are confirming the
superiority of the proposed method. Figs.2,3&4 show the
effectiveness of the proposed method over the standard image
contrast enhancement and resolution enhancement techniques.
Different images with different features are used for comparison.
From Fig.2 we see that input image is grey scale image with size
[256256] and the resultant image of size [512512] with sharp
details and with high contrast image.Fig.3 represents the
IJSET@2016

ISSN:2277-1581
01 Nov. 2016

proposed method for color image. Fig .3(d) shows the output of
CLAHE method. However in image details are over enhanced.
Fig .3(e) shows the resultant of proposed method. On comparing
with 3(d) proposed method resultant mitigate the contrast over
stretching.
Peak signal to noise ratio (PSNR): It is the ratio between the
maximum possible signal powers to the noise power. Here noise
means mean of square of error between input image and
enhanced image and it is called as mean square error (MSE), it is
defined as follows

------(2)
Where M, N are the height and weight of the input image .K is
the output enhanced image. The peak signal to noise ratio for
gray scale image is given
-------(3)
Since color images have three deferent channels for color
representation, mean square error (MSE) for color images is
given
as
follows[10]

M, N are the dimensions of the image

represents R channel
color component of the input image at pixel location (i ,j) and
represents R channel color component of the output image
at pixel location (i ,j) .
NOISE ESTIMATION (NE): It is used to measure the noise
amplification during the process of resolution enhancement and
contrast enhancement. For the estimation of noise ,noise is
added to input image with mean zero and with a variance of
2.56. Table 1 gives the psnr , rmse , standard deviation values of
gray scale images and color images individually.
V . CONCLUSION
In this paper we present a novel image resolution
enhancement and contrast enhancement methods, which
combine the DWT-SWT with CLAHE-SWT. In this method the
image is decomposed into sub-bands by both DWT and SWT.
The high frequency sub-bands of DWT are interpolated and
combined with high frequency sub-bands of SWT. This increases
the resolution of image .CLAHE is applied to only lowfrequency sub-bands of SWT. This reduces the effect of noise.
Visual results and quantitative results (PSNR, NE, RMSE) show
that proposed method performs better than other existing
methods.

doi : 10.17950/ijset/v5s11/1103

Page 253

International Journal of Scientific Engineering and Technology

Volume No.5 Issue No.6, pp: 507-511

ISSN:2277-1581
01 Nov. 2016

GREY SCALE IMAGES:

(b)[512512]

(a)[256256]

(c) [256256]

COLOUR IMAGES:

(d) [256256]

COLOR IMAGES :

(d)[256256]

(c)[256256]

(e) [256256]

(f) [512512]
Fig 3. CAPS Image. a) ORIGINAL IMAGE (b) DWT-SWT
c) HISTOGRAM EQUALIZATION d) CLAHE e) CLAHEDWT and f) PROPOSED METHOD

(e)[256256]

(f)[512512]
(a) [256256]

(b) [512512]

(c) [256256]

(d) [256256]

(e) [256256]

(f) [512512]

Fig 2. Gray Scale Image. a) ORIGINAL IMAGE (b) DWT-SWT

c) HISTOGRAM EQUALIZATION d) CLAHE e) CLAHE-DWT
and f) PROPOSED METHOD

COLOUR IMAGES:

Table 1. Comparison of Proposed method with other

methods using Average values of all 3 image
ON
AVERAGE
HE

PSNR(db.)

NE()

RMSE

53.75

12.73

17.79

CLAHE

52.45

16.00

17.24

[256256]

[512512]

CLAHEDWT
DWT

55.59

10.43

11.77

36.2263

32.38

20.74

DWT&SWT

43.0733

19.55

PROPOSED
METHOD

58.24

04.50

IJSET@2016

Fig 4. WALL Image. a) ORIGINAL IMAGE (b) DWT-SWT c)

HISTOGRAM EQUALIZATION d) CLAHE e) CLAHE-DWT and
f) PROPOSED METHOD

14.33

doi : 10.17950/ijset/v5s11/1103
05.30

Page 254

International Journal of Scientific Engineering and Technology

Volume No.5 Issue No.6, pp: 507-511

ISSN:2277-1581
01 Nov. 2016

and discrete wavelet transform for image enhancement. The institution

of engineering and technology,2015 vol 9,issue10,pp 908-915
ix.
VijeeshGovind, Arun A. Balakrishnan , Dominic Mathew et
al.:A novel approach for contrast enhancement and noise removal of
medical images. Control Communication and Computing (ICCC), 2013
International Conference
x.
Henrykpalus ,mariusz .Color Image Quantization using and
KHM Clustering Techniques with Outlier-Based Initialization .July 16
2015,Pages 261-278.
xi.
Vijay A. Kotkar and Sanjay S. Gharde.Imagecontrst
enhancement by preserving the brightness using local and global
features.3rd International conference on computational Intelligence and
Informational technology, Mumbai..

T. V. Hyma Lakshmi obtained her

M.Tech.(ECE).,
from
JNTUCE,
Ananthapur and B.E. (ECE)from
S.R.K.R,
Bhimavaram.
Presently
working as Assistant Professor in
S.R.K.R.Engineering
college
Bhimavaram and pursuing Ph.D.
Program from K L University.

Fig. Comparison of PSNR, Noise Estimation, RMSE of

existing methods vs. proposed method.
I . References:
i.
Hassan Demirel and Gholamreza Anbarjafari. Discrete
Wavelet Transform-Based Satellite Image Resolution
ii.
Enhancement IEEE transaction on geosciences and remote
sensing, vol. 49, NO. 6, june 2011.
iii.
Hassan Demirel and Gholamreza Anbarjafari. Image
resolution enhancement by using discrete and stationary wavelet
decomposition .IEEE transactions on image processing may 2011.
iv.
Abdulla-al-Wadud,Md Hassanul Kabir A Dynamic histogram
equalization for image contrast enhancement IEEE transaction on
consumer electronics , Volume:53, issue 2, 2007.
v.
S. M. Pizer, E. P. Amburn, J. D. Austin, et al.: Adaptive
Histogram Equalization and Its Variations. Computer Vision, Graphics,
and Image Processing 39 (1987) 355-368..
vi.
). Byong Seok Min, Dong Kyun Lim, Seung Jong Kim and
Joo Heung Lee et al.: Anovel method of determining CLAHE based on
image entropy. International Journal of Software Engineering and Its
Applications,Vol.7, No.5 (2013), pp.113-120.
vii.
C. Wang and Z. Ye, Brightness Preserving Histogram
Equalization with Maximum Entropy: A Variational Perspective,
IEEE Trans. on Consumer Electronics, vol. 51, no. 4, (2005), pp. 13261334.
viii.
Huang
lidong,
zhaowei,
wangjun,
sun
zebin
,etal.:combination of contrast limited adaptive histogram equalization

IJSET@2016

Dr.Tenneti Madhu obtained his B.E. degree

from University of Madras, M.Tech from
REC, Kurukshetra in 1994 and PhD from
Osmania University in 2004.His research
interests include GPS Data Analysis, Image
Processing, Nano Technology and VLSI
design. Presently working as Principal in
S.I.E.T, Narsapur.

doi : 10.17950/ijset/v5s11/1103

Shaik. Esub basha currently studying

M.Tech In S.R.K.R engineering
college, bhimavaram. And completed
his Bachelor Degree in SMCE
engineering college, tummalapalem,
Guntur district, Andhra Pradesh

Dr. K. Ch. Sri Kavya received Doctorate degree

from Koneru Lakshmaiah Education Foundation
(KL University) for her research work on Array
Antennas in 2014. She received her Bachelors
in ECE and Masters in Communication and
Radars systems, from KLCE affiliated to
Acharya Nagarjuna University in 2003 and in
2008 respectively. She has been currently
working as a Professor in the Department of
Electronics and Communication Engineering in
KL University.

Page 255