Content uploaded by Anggunmeka Luhur Prasasti
Author content
All content in this area was uploaded by Anggunmeka Luhur Prasasti on Aug 27, 2019
Content may be subject to copyright.
International Journal of Applied Engineering Research ISSN 0973-4562 Volume 12, Number 23 (2017) pp. 13596-13603
© Research India Publications. http://www.ripublication.com
13596
A Review of Image Enhancement Methods
Ridho Dwisyah Putra
Student, Computer Engineering,
Faculty of Electrical Engineering, Telkom University, Bandung, Indonesia.
Tito Waluyo Purboyo
Lecturer, Computer Engineering,
Faculty of Electrical Engineering, Telkom University, Bandung, Indonesia.
Anggunmeka Luhur Prasasti
Lecturer, Computer Engineering,
Faculty of Electrical Engineering, Telkom University, Bandung, Indonesia.
Abstract
Today, face recognition is a technology that has been widely
applied to various purposes, especially related to the image.
Examples of the use of face recognition include face
recognition for the manufacture of identity cards as well as
driving identity cards. In addition, it is also widely used in a
security system somewhere that can recognize the face of
someone. At this time, face recognition has many variants.
However, there are still some drawbacks due to the image
capture technique or not optimal yet. This can be caused by
poor image retrieval, insufficient shooting distance, and
minimal lighting conditions. One solution that can be done to
overcome the drawback to the image taken is by using the
method of image enhancement. One method of image
enhancement that can be used is image enhancement.
Keywords: Face Recognition, Image Enhancement, Matlab,
Histogram Equalization, Median Filter.
INTRODUCTION
The use of the camera as a tool that enabled to take pictures has
been very much used by the community today. Cameras are
widely used for a variety of needs ranging from such uses to
support security systems by identifying people caught by the
camera. One of the information that can be used as
identification is part of a human face. The face of a person has
a unique characteristic that is different in each person.
Face recognition is a method where in a face represents a 3D
image in which there are levels of brightness, lighting, pose,
expression and others which then made the identification
process to its 2D image. Face recognition sometimes can’t be
implemented perfectly so that the processed image can’t be
recognized properly. This makes the necessary process of
image enhancement in order to obtain a better image quality
again.
Face recognition techniques at this time even growing rapidly
with the emergence of variants of face recognition methods. By
using face recognition it can be done to identify someone based
on the identity and characteristic of the person's face based on
the image of the caught from the camera used.
But to note is that sometimes there are some images obtained
do not have a good visual form or still have noise. This can be
caused by several factors, primarily due to the minimal level of
illumination that can cause the image color to be different from
the original color of the object. This is known as color
constancy.
Color constancy is a color determination possessed by humans
where the colors received by the human vision system still have
a relatively constant color even though the object is in a place
that has minimal lighting. To overcome this problem is used
image enhancement method to improve the quality of the
image.
Image enhancement is one of the initial processes in image
processing. Image enhancement serves to improve the quality
of the existing image. Some of the processes included in the
image enhancement section include image brightness change,
contrast enhancement, contrast stretching, image histogram
conversion, image softening, sharpening, edge detection,
histogram equalization and geometric alteration.
LITERATURE REVIEW
Point Operation
a. Brightness Adjustment
Image improvement method by performing brightness
adjustment is one of the methods of image repair which is quite
International Journal of Applied Engineering Research ISSN 0973-4562 Volume 12, Number 23 (2017) pp. 13596-13603
© Research India Publications. http://www.ripublication.com
13597
simple. In general, each pixel is concentrated on one side of the
histogram using a certain gray level range.
If the image brightness value is increased then the image
histogram will be shifted to the right. If the brightness value is
lowered then the image histogram will shift to the left which
then will impact on the brightness level of the image.
Mathematically, brightness adjustment operations can be
defined as follows:
Where is an image after using the brightness
adjustment method and is the image before the process
of brightness adjustment.
Where b is the brightness intensity value used to make image
brightness adjustment. If b has a positive operand then the
image will be brighter and if b has negative operand then the
image brightness level will decrease or dim.
It can be seen, that the process of brightness adjustment is done
by adding the value of each pixel with a certain constant.
If on the grayscale image pixel value after the brightness
adjustment is more than the maximum intensity value that can
be accommodated is 255 then the pixel value will be made 255.
If the pixel the brightness adjustment is less than the minimum
value that has been set that is equal to 0 then the pixel value
will be made 0
b. Gamma Correction
Gamma refers to the brightness of the image. Gamma self-
regulation is used to determine the dim light of the displayed
image. The purpose of the gamma correction is the same as the
brightness adjustment method that improves the image from the
illumination side. However, the brightness adjustment has a
linear arrangement.
In the gamma correction, the arrangement is done by utilizing
certain functions on each input that will determine the output
of the image. When depicted in graphical form, the function of
the gamma is curve-shaped. The darkest and brightest areas of
the gamma graph will not have much effect on different gamma
arrangements. However, the middle area of the graph will have
an effect by following the arrangement. The equations of
gamma can be defined as follows:
Where is the image of the gamma correction process
and is the image before the gamma correction process.
The symbol is a gammon correction factor with a value range
of 0 < <1. If γ has a value less or less than 1 to near 0 then the
resulting image will be brighter. If is equal to 1 then the
resulting image will be similar to the input image used.
The use of gamma correction because the input signal provided
is not enough to be able to display a good image. Thus, if no
gamma adjustment is made then the resulting image will be
difficult to see. The gamma correction method will produce a
brighter and more natural image.
c. Contrast Stretching
Contrast stretching is represented in light and dark distribution
in an image. Contrast itself can be divided into three types as
follows:
1. Low contrast. Images that have low contrast are mostly
bright or partially dark. Each pixel in the image will be
concentrated in one region either in the left, right or middle
regions of the histogram.
When the pixel values converge on the left the image looks
darker. If the pixel values are clustered on the right then
the image tends to be lighter. If the image pixel is
assembled in the middle of the histogram then the resulting
image is neither too bright nor too dark.
2. Good contrast. In the image with good contrast, the gray
degree range is owned flat so that no one dominates. So
that the histogram can be seen that there is no maximum
point or minimum point.
3. High contrast. In high contrast, the width of the gray width
is similar too good contrast. However, in the high contrast
there are two colors that dominate the dark and light. Thus,
in the resulting histogram there are two peaks is peaks with
the low gray degree and peaks with the high gray degree.
d. Histogram Equalization
Histogram equalization is a method of contrast adjustment by
using the histogram of the image. This histogram can improve
the quality of the image. The workings of this histogram are to
increase the peak of the histogram and decrease the minimum
histogram point. This is done so that the spread of the value of
each pixel can be done evenly or not much different.
Histogram equalization causes the dynamic range to stretch
with the density distribution of the image made homogeneous
so that the image contrast can be increased. The histogram
equalization can be defined if the histogram of I(x, y) contained
in a pixel with the gray level is i (where i is 0, 1, ..., k-1) and
is the number of pixels at I(x, y) by the gray level is i.
Where is the number of pixels that have the gray degree and
n is the total number of pixels in an image. In the above
equation, it is stated that a mapping of any original pixel
intensity value is 0 to 255.
It is also assumed that the equalization histogram changes the
input value of rk to sk which then changes the value from s to
be defined as follows :
International Journal of Applied Engineering Research ISSN 0973-4562 Volume 12, Number 23 (2017) pp. 13596-13603
© Research India Publications. http://www.ripublication.com
13598
Where, L is defined as the gray level, is the number of pixel
frequencies that appear in the image and sk defined as the gray
intensity value of a pixel in the image.
Read input im age
Histogram equalization
function
Class 8 bits image
Show input an d output
histograms and images
Figure 3: Histogram Equalization Flowchart
In figure 3, we can see about histogram equalization flowchart.
For the first step, we have to read input image. Then, we use
histogram equalization function with 8 bits from the image. The
final steps show the output image and output histogram.
After that, the next process is to compare between the input
image with the image of the results that have been using the
histogram equalization image enhancement method. So, we can
see the difference between the image before enhancement with
image enhancement using histogram equalization.
e. Contrast Limited Adaptive Histogram Equalization
Basically, the contrast limited adaptive histogram equalization
method has the same principle as the histogram equalization
method. In this method, the image is divided into several sub-
images with size n x n. after that, just done histogram
equalization process in accordance with sub-image division
that has been done before.
Contrast limited adaptive histogram equalization improves
image contrast by changing the intensity value of the image.
This method operates on a smaller pixel area range when
compared to the histogram equalization method.
Each image pixel contrast is enhanced so that the output image
histogram is compatible with the specified image. Using
neighboring pixels is then merged using interpolation to
remove the existing boundary. The contrast on a homogeneous
region is restricted. This is done to avoid the appearance of
noise in the image.
The steps in the contrast limited adaptive histogram
equalization method are described as follows [1]:
a. The input image used is divided into several sub-image.
b. Then, from each sub-image that has been divided the
calculation of the histogram for each sub-image.
c. Histogram results from sub-image in a clip. The pixel
values in the sub-image are evenly distributed on each
grade of gray in the image.
The average for each number of pixels at each degree of gray
is defined as follows:
Where is expressed as the average of the number of pixels
available. The number of pixels with x coordinates in the sub-
image is represented by symbol. represents the
number of pixels in y coordinates in the sub-image. While
is the number of gray degree values in the image.
Based on the above equation, the value of the clip-limit is
defined as follows:
Where is the actual value of the clip-limit and is the
maximum value of the average pixel on each grade of gray in
the image.
In the input image histogram, a clip is performed if the total
pixel is greater than the total value. The number of pixels
is distributed evenly into each grade of gray degree which can
be defined as follows:
Where is the number of distributed pixels and is the
total pixel that is clipped. To calculate the distribution of the
histogram that has been divided into several sub-images in the
initial process we can define the equation as follows:
+
Where is a pixel found at the gray level in each sub-image
and i is the value of the gray degree of the image. After
calculating the histogram distribution in the sub-image the next
International Journal of Applied Engineering Research ISSN 0973-4562 Volume 12, Number 23 (2017) pp. 13596-13603
© Research India Publications. http://www.ripublication.com
13599
process is to do the pixel distribution defined by the equation
as follows:
Where S is the result of the pixel distribution. While is the
number of pixels that have been cut.
The use of contrast limited adaptive histogram equalization
method will keep track of all pixel values from the minimum to
the maximum.
Thus, this method will distribute each pixel in the image. If the
tracking process stops before all pixels are distributed evenly
then the process will be repeated starting from equation 9 until
all pixels are disbursed separately and a new image histogram
is obtained.
d. The limited contrast histogram of each sub-image is
processed using histogram equalization and mapped using
linear interpolation.
Spatial Filter
a. Linear Filter
1. Mean Filter :
Mean filter is used to smoothing the image by calculating the
average value of pixels in the image. Mean filter is included in
the type of spatial filter because in the process include the
pixels around it.
In the process, mean filtering involves surrounding pixels. The
pixels to be processed are included in a matrix of dimension N
x N. Mathematically, the mean filtering has the same weight as
the neighboring pixel defined as follows:
(10)
Where represents the image of the result. while
represents the input image used. The upper bound value of m
and n represent the size of the row and column of the mean
filtering. If the mean filtering does not have the same weight as
the neighboring pixels, then the process is using convolution.
2. Gaussian Filter :
The gaussian filter is included in a linear type filter with a
weight value for each pixel set in it by using the gaussian
function. The gaussian filter is widely used for smoothing,
blurring and eliminating noise in the image.
The linear process in the gaussian filter is done by multiplying
each adjacent neighbor pixel and summing the result so that it
gets the result for a certain coordinate point denoted as (x, y).
The mechanism of the linear spatial filter is to move the center
of a filter mask from one point to another. In each pixel (x, y),
the result of the filter at that point is the sum of the
multiplication of the filter coefficients and the corresponding
neighbor pixels in the filter mask range.
There are two components to note on the gaussian filter that is
correlation and convolution. Correlation is the process of
passing mask to the image. While the definition is defined as a
process for obtaining pixel values based on their own pixel
values, neighboring pixels and kernel matrices.
In the process, the kernel will be shifted along the rows and
columns of the input image used so that the new pixel value of
the resulting image will be obtained. In the gaussian filter itself,
the convolution process first rotates the filter mask of 180o and
then passes the image.
Gaussian filters have two types of filters: one-dimensional
gaussian filter and two-dimensional gaussian filter. The one-
dimensional gaussian function is defined as follows:
Where is expressed as the standard deviation of the
distribution. As the value of gets larger then the distribution
curve of the gaussian gets wider and the peak decreases.
The two-dimensional gaussian form is defined as follows:
Where σ is expressed as the standard deviation of the same
distribution as the one-dimensional gaussian function. For x
and y are expressed as coordinate points (rows and columns) in
image pixels.
b. Non-Linear Filter
1. Median Filter :
The median filter is image enhancement method to reduce noise
in the image. The median filter uses a way to extract certain
data sections of a set of data, by eliminating the parts of
undesirable data. There are several types of filter used in image
processing one of which is a spacial filter.
Spatial filters are also called discrete convolution filters or
filters that convolute an image with another image. The size of
this image filter is usually small, relative to the image and is
called convolution mask. This operation copies an image on a
pixel resulting in a different effect. With spatial filters, the
computations performed will only result in the value of the
neighboring pixels and pixels.
International Journal of Applied Engineering Research ISSN 0973-4562 Volume 12, Number 23 (2017) pp. 13596-13603
© Research India Publications. http://www.ripublication.com
13600
In spatial frequency, the value of pixels is combined to form a
single pixel value. The use of spatial filter techniques in the
image, generally the point to be processed along with the points
around it is inserted into a matrix of size N x N. This matrix is
called matrix neighbor (neighboring matrix), where N is large
depending on the need, but generally N this is always an odd
multiplier because the point to be processed is placed in the
middle of the matrix.
In addition to the use of the neighboring matrix, spatial filter
techniques use a matrix that is matrix convolution (mask) the
same size as the neighboring matrix.
One type of filter included in the spatial filter is the Median
Filter. The median filter works by evaluating the brightness
level of a pixel and determining which pixels whose brightness
is the median value of all pixels.
The median value is determined and puts the pixel brightness
on the stratified order and selects the middle value. so that the
number obtained from the existing pixel brightness becomes
less than and more than the middle value obtained. The median
filter will eliminate image pixels that differ considerably with
another image.
2. Conservative Filter:
Conservative filter is one technique to reduce existing noise on
the image. In the median filter, the filter process uses the middle
value of the processed neighboring pixels and pixels. In the
conservative filter, the values used are the minimum and
maximum values but excluding the processed middle pixels.
In the conservative filter, the calculation process is performed
if the middle pixel is in the range between the minimum and
maximum values, the new center pixel value will remain the
same as the pre-existing value.
If the middle pixel is greater than the maximum value that
exists in the surrounding pixels then the middle pixel value is
replaced with the maximum value. If the middle pixel is smaller
than the minimum pixel value around it, the middle pixel value
will be equal to the minimum value.
Retinex
Retinex is used to improve the quality of digital images that
have a relationship with the quality of lighting while
maintaining color constancy. Color constancy is the firmness
of the object that has the perception that the color of the object
is felt to remain constant in various lighting conditions.
When the dynamic range of an image exceeds the dynamic
range of the medium, the visualization of color and detail
appears to be weaker than the actual image. Color constancy
aims to produce color to look the same as the difference in
vision and lighting conditions.
Image Smoothi ng
Estimation Normalization
Input Image Log
Log
-
+
Normalize Imag e
Figure 2: Retinex Diagram
In figure 2, as we can see above that the retinex method is
divided into two stages. The first stage of illumination in the
estimated input image is smoothed. Then the second stage is
normalized by using logarithmic differences between the input
image and illumination estimation.
At this time, retinex itself experienced a lot of development.
Some of the development variants of retinex are Single Scale
Retinex (SSR) and Multi-Scale Retinex (MSR).
Single Scale Retinex can be defined by the equation as follows:
Where is the output of retinex and represent
the image distribution in three spectral bands is spectral bands
for red, green and blue. In the above equation, the symbol
represents the convolution. As for F (x, y) represents the
gaussian function. The gaussian function is defined as follows:
Where is the gaussian constant which is usually used as the
standard deviation and r is defined as follows :
Where r defined as a gaussian constant. Small gaussian
constants can provide good dynamic range compression. But,
the color produced by the image is not very good. However, on
a large scale the color will be better. For the value of K is
defined as follows:
Where, the above equation is used as the definition of the value
of K. Up to this point, Single Scale Retinex has been able to
perform dynamic range compression on the image so that on a
low scale can be done image reinforcement in the dark and
weaken the bright image. On a large scale, Single Scale Retinex
is also capable of producing substantial image brightness and
produce a more natural image.
International Journal of Applied Engineering Research ISSN 0973-4562 Volume 12, Number 23 (2017) pp. 13596-13603
© Research India Publications. http://www.ripublication.com
13601
REVIEW AND DISCUSSIONS
No.
Image Enhancement
Methods
Description
1
Brightness Adjustment
Brightens the image by raising the image brightness level. The image histogram
will shift to the right when the brightness level is raised and the image is brighter.
If the brightness level is lowered then the image histogram will shift to the left and
the resulting image will darken.
2
Gamma Correction
Gamma correction is also used to improve the brightness of the image. It is the
same as the brightness adjustment method. The difference with brightness
adjustment is gamma correction method using correction factor which has the
range value 0 <γ <1. The higher the correction factor is, the image will be more
similar to the input image. The smaller the correction factor the lighter the
resulting image.
3
Contrast Stretching
Contrast stretching is done by distributing bright and dark pixels to the image. The
use of contrast stretching method is very good used in low contrast image. This
method works well on images that have gaussian distributions.
4
Histogram Equalization
Histogram equalization has a function in improving the image by distributing the
intensity of the histogram to appear more evenly although it cannot be uniformed
completely.
5
Contrast Limited
Adaptive Histogram
Equalization
This method is similar to histogram equalization method. However, the difference
is in sub-image division. Each image is divided into several sub-images. Then,
every block that has been divided is done histogram equalization process.
6
Mean Filter
The mean filter method serves to reduce the pixel intensity value with the next
pixel. The mean filter calculates the value for the new pixel with the average value
of the neighboring pixel and pixel in question.
7
Gaussian Filter
The gaussian filter is a filter that works to smoothing, blurring and eliminating
noise in the image.
8
Median Filter
The median filter is included in a non-linear filter because it does not use a
convolution process. The calculation process is done by sorting the intensity value
of each pixel. Then, the original pixels will be replaced with new pixels that have
been through the counting process. The function of this filter to smooth the image
and reduce noise.
9
Conservative Filter
Conservative filters use the minimum point and maximum point of neighboring
pixels.
10
Retinex
The retinex is a combination of words from the retina and cortex. The retinex
method is used to improve image quality from the illumination side obtained by
the image while maintaining color constancy. Color constancy is the firmness of
color perception on an object. Thus, the color of an object is relatively fixed under
various lighting conditions.
CONCLUSION AND FUTURE WORKS
This paper provides an overview of the concept of image
processing. This paper is discussed from the various literature
on image improvement methods. Image enhancement
methods are needed to help the quality of the image better.
Thus, the resulting image gets a decent quality to be seen by
the human vision system. In addition, by using the method of
image enhancement can be seen the difference between the
image before using image enhancement with the image after
using the image enhancement methods.
International Journal of Applied Engineering Research ISSN 0973-4562 Volume 12, Number 23 (2017) pp. 13596-13603
© Research India Publications. http://www.ripublication.com
13602
In future research, it is necessary to discuss and research on
other image improvement methods. Image enhancement
method that need to be developed in research is about of
frequency domain methods and hybrid methods to improve
image quality better. There are still many things that can be
developed in improving image quality with various methods.
ACKNOWLEDGEMENT
I as the author of this paper would like to say a big thank you
to Telkom University for the various support that has been
given to complete the research that has been implemented.
REFERENCES
[1] Zulkifli, “Teknik Pengolahan Citra Menggunakan
Metode Kecerahan Citra Kontras dan Penajaman Citra
dalam Menghasilkan Kualitas Gambar”, Jurnal
Variasi, Volume 08, No. 01, Desember, 2016.
[2] H. Singh, N. Agrawal, A. Kumar, G.K. Singh, H. -N.
Lee, “A Novel Gamma Correction Approach Using
Optimally Clipped Sub-Equalization for Dark Image
Enhancement”, IEEE, 2016.
[3] Hany Farid, “Blind Inverse Gamma Correction”, IEEE
Transactions on Image Processing, Vol. 10, No. 10,
October, 2001.
[4] Chiaki Ueda, Tadahiro Azetsu, Noriaki Suetake, Eiji
Uchino, “Gamma Correction-based Image
Enhancement for Elderly Vision”, 15th International
Symposium on Communications and Information
Technologies (ISCIT), 2015.
[5] Manjot Kaur Khehra, Mandeep Devgun, “Survey on
Image Enhancement Techniques for Digital Images”,
Scholars Journal of Engineering and Technology
(SJET), 2015.
[6] Nur Wakhidah, “Perbaikan Kualitas Citra
Menggunakan Metode Contrast Stretching”, Jurnal
Transformatika, Volume 8, No. 2, Januari, 2011.
[7] Shanto Rahman, Md. Mostafijur Rahman, Khalid
Hussain, Shah Mostafa Khaled, Mohammad Shoyaib,
“Image Enhancement in Spatial Domain: A
Comprehensive Study”, 17th International Conference
on Computer and Information Technology (ICCIT)),
2014.
[8] Youlian Zhu, Cheng Huang, “An Adaptive Histogram
Equalization Algorithm on the Image Gray Level
Mapping”, International Conference on Solid State
Devices and Materials Science, Physics Procedia 25
601-608, 2012.
[9] Budi Hartono, Veronica Lusiana, “Analisa Teknik
Adaptive Histogram Equalization dan Contrast
Stretching untuk Perbaikan Kualitas Citra”, Jurnal
Teknologi Informasi DINAMIK, Volume 19, No. 1,
Januari, 2014.
[10] Dinesh Sonker, M.P. Parsai, “Comparison of
Histogram Equalization Techniques for Image
Enhancement of Grayscale Images of Dawn and
Dusk”, International Journal of Modern Engineering
Research (IJMER), Vol. 3, Issue. 4, Jul-Aug, 2013.
[11] Sayali Nimkar, Sucheta Shrivastata, Sanal Varghese,
“Contrast Enhancement and Brightness Preservation
Using Multi-decomposition Histogram Equalization”,
Signal & Image Processing : An International Journal
(SIPIJ), Vol. 4, No. 3, June, 2013.
[12] Rajesh Garg, Bhawna Mittal, Sheetal Garg,
“Histogram Equalization Techniques for Image
Enhancement”, International Journal of Electronics &
Communication Technology (IJECT), Vol. , Issue 1,
March, 2011.
[13] Sunil Malviya, Hemant Amhia, “Image Enhancement
Using Improved Mean Filter at Low and High Noise
Density”, International Journal of Emerging
Engineering Research and Technology, Volume 2,
Issue 3, June, 2014.
[14] Nafis Uddin Khan, K.V. Arya, Manisha Pattanaik,
“An Efficient Image Noise Removal and Enhancement
Method”, IEEE, 2010.
[15] Sandeep Singh, Sandeep Sharma, “A Survey of Image
Enhancement Techniques”, IPASJ International
Journal of Computer Science (IIJCS), Volume 2, Issue
5, May, 2014.
[16] Xu Guan, Su Jian, Pan Hongda, Zhang Zhiguo, Gong
Haibin, “An Image Enhancement Method Based on
Gamma Correction”, Second International Symposium
on Computational Intelligence and Design, 2009.
[17] V. Janani, M. Dinakaran, “Infrared Image
Enhancement Techniques – A Review”, 2nd
International Conference on Current Trends in
Engineering and Technology (ICCTET), Coimbatore,
India, July 8, 2014.
[18] Q. Mirsharif, F. Tajeripour, “Investigating Image
Enhancement Methods for Better Classification of
Retinal Blood Vessels Into Arteries and Veins”, The
16th CSI International Symposium on Artificial
Intelligence and Signal Processing (AISP), 2012.
[19] Neethu Lekshmi J M, Shinny. C, “Color Image
Enhancement Using Retinex Algorithm”, National
Conference in Emerging Technologies (NCET), IOSR
Journal of Computer Engineering (IOSR-JCE), 2014.
International Journal of Applied Engineering Research ISSN 0973-4562 Volume 12, Number 23 (2017) pp. 13596-13603
© Research India Publications. http://www.ripublication.com
13603
[20] Di Li, Yadi Zhang, Pengcheng Wen, Linting Bai, “A
Retinex Algorithm for Image Enhancement Based on
Recursive Bilateral Filtering”, 11th International
Conference on Computational Intelligence and
Security, 2015.
[21] G. Hemalatha, C.P. Sumathi, “Prepocessing
Techniques of Facial Image with Median and Gabor
Filters”, Interantional Conference on Information
Communication and Embedded System (ICICES),
2016.
[22] Ying Li, Changzhi Hou, Fu Tian, Hongli Yu, Lei Guo,
Guizhi Xu, Xueqin Shen, Weili Yan, “Enhancement of
Infrared Image Based on the Retinex Theory”,
Proceedings of the 29th Annual International
Conference of the IEEE EMBS, Lyon, France, August
23-26, 2007.
[23] Bambang Yuwono, “Image Smoothing Menggunakan
Mean Filtering, Median Filtering, Modus Filtering dan
Gaussian Filtering”, Telematika, Vol. 7, No. 1, July,
2010.
[24] Sin Hoong Teoh, Haidi Ibrahim, “Median Filtering
Frameworks for Reducing Impulse Noise from
Grayscale Digital Images: A Literature Survey”,
International Journal of Future Computing and
Communication, Vol. 1, No. 4, December, 2012.
[25] Bo Li, Shuhang Wang, Yanbing Geng, “Image
Enhancement Based on Retinex and Lightness
Decomposition”, 18th IEEE International Conference
on Image Processing, 2011.
[26] R. Lenka, Dr. A. Khandual, “A Study on Retinex
Theory and Illumination Effects”, International
Journal of Advanced Research in Computer Science
and Software Engineering, Volume 6, Issue 1,
January, 2016.
[27] Doo Hyun Choi, Ick Hoon Jang, Mi Hye Kim, Nam
Chul Kim, “Color Image Enhancement Using Single-
Scale Retinex Based on an Improved Image Formation
Model”, 16th European Signal Processing Conference
(EUSIPCO), Lausanne, Switzerland, August 25-29,
2008.
[28] C.M. Sujatha, K. Navinkumar, K. S. Arunlal, S. A.
Hari Prasad, “Performance Evaluation of
Homomorphic Filtering, Anisotrophic Filtering and
Autocontrast Algorithm”, International Conference
on Advances in Computing, Control, and
Telecommunication Technologies, 2009.
[29] Sunil Kumar Kopparapu, M Satish, “Identifying
Optimal Gaussian Filter for Gaussian Noise
Removal”, Third National Conference on Computer
Vision, Pattern Recognition, Image Processing and
Graphics, 2011.
[30] Ankit Aggarwal, R. S Chauhan, Kamaljeet Kaur, “An
Adaptive Image Enhancement Technique Preserving
Brightness Level Using Gamma Correction”, Advance
in Electronic and Electric Engineering, Volume 3,
Number 9, 2013.