Iisrt zzz bhavyasri vanteddu

International Journal of Electrical and Computing Engineering (IJECE)
Vol. 1, Issue. 4, June 2015 ISSN (Online): 2349-8218
59
Abstract: This paper presents an algorithm for Image
enhancement. This algorithm deals with the Bright-
pass filter (BPF) which decomposes an image into
Illumination and Reflectance. This BPF is
implemented with and without using Block
processing. The Bi-log transformation is applied on
the illumination component andthen it is synthesized
with reflectance which results in the final enhanced
image. Both the results of this algorithm with and
without using Block processing in comparison with
Brightness Preserving Dynamic Histogram
Equalization (BPDHE) algorithm are presented.
Index Terms: Image Enhancement, Bi-log
transformation, Block processing.
I. INTRODUCTION
The principle objective of Image
enhancement is to process an image so that the
result is more suitable than the original image for
specific applications. Up to now, Image
enhancement has been applied to varied areas of
science and engineering, such as Atmospheric
sciences, Astrophotography, Biomedicine,
Computer vision, etc.
The enhancement methods can broadly be
divided in to the following two categories. They are
Spatial domain methods, which operate directly on
pixels. Frequency domain methods, which operate
on the Fourier transform of an image.
A literature survey [1-16] on Image
enhancement techniques is done by referring to the
research work done during 2000-2014. The several
number of methods are available in image
enhancement such as processing an image by
decomposing the image based on specific criteria,
using the filter or a group of filters, modifying the
multi-scale measure, constructing HDR image with
multiple images, using different transformation
methods, etc.
The image enhancement techniques differ
by their features. Some of them are enhancing the
low contrast or medium contrast details, images
whose contrast details vary across the image, only a
part of the image or different parts of the image to
different extents, simultaneously both contrast and
sharpness of the image and integrating the
luminance and contrast masking phenomena.
The Block processing function helps to
process the image individually on each block by
dividing the original image into rectangular blocks
of required size specified by the two-element
block-size vector. After processing, the results are
assembled into an output image. In this algorithm,
the block-size is taken as 15*15.
II. BRIGHT-PASS FILTER
Many algorithms result in over-
enhancement since they do not consider the range
of reflectance. Hence the BPF [15] was considered.
This filter is mainly used to restrict the range of
reflectance to [0], [1]. In the BPF, the effect on the
central pixel of value b caused by an adjacent pixel
of value a is positively related to the frequency for
pixels of value a and pixels of value b being
neighbours all over the image.
In general, the neighbours can be defined
flexibly for different applications. In the BPF, the
weight of adjacent pixels is considered as its
normalized version as the frequency is static. The
neighbours are considered since it was already
known that there is no obvious difference between
the filtering results by using slightly different
neighbours. The filtering results obtained using
four-connectivity and eight-connectivity are
similar. For ease, the neighbours of a pixel G(x, y)
in four connectivity was given as:
NB(x, y) = {G(x, y − 1), G(x, y + 1), G(x − 1, y),
G(x + 1, y), G(x, y)} (1)
The frequency (k, l) for pixels of values
k and l to be neighbours all over the image is given
as:
( ) ∑ ∑ ( )
(2)
where NNk,l(x, y) indicates the number of its
neighbours of value l, m and n are the height and
the width of the image.
Image Enhancement using Bright Pass Filter (BPF)
1
V.Bhavyasri, 2
N.Udaya Kumar, 3
K.Bala Sindhuri
1
PG Student, SRKR Engineering College, Bhimavaram, INDIA
2
Professor, SRKR Engineering College, Bhimavaram, INDIA
3
Assistant Professor, SRKR Engineering College, Bhimavaram, INDIA

60
The frequency signal ( ) suffers from
noise and varies roughly. The local mean Q(k, l) of
the frequency signal is considered:
( ) (∑ ( ))/(2.win+1)
(3)
where win is the window size. The size of the
window should not be too small or too large, to
remove the noise as well as preserve the local trend
of the frequency. The range of the gray levels was
used to set the size of the window. The window
size is set as follows:
win = (max(G(x, y)) − min(G(x, y)))/32 (4)
The bright-pass filter, BPF[·], is defined
as the weighted average of adjacent pixels with the
weight positively related to the frequency Q(k, l)
and is evaluated as follows:
BPF[G(x,y)]=
( )
∑ ( ( ( ) ( )) ( ( ) ( ) ( )))( )
(5)
where Ω denotes the local patch of size 15*15
centered at coordinate (x, y). To ensure that only
brighter neighbours are taken into account, the unit
step function U(x, y) is utilized and W(x, y), the
normalization factor ensures the sum of pixel
weight to be 1.
W(x,y)=∑ ( ( ( ) ( )) ( ( ) ( )))( ) (6)
Image Decomposition:
BPF is used here to evaluate the
illumination. The reflex lightness is termed as the
product of reflectance and illumination according
to the Retinex theory. Illumination means light cast
on the surface of the scene. Reflectance means
surface reflects more than light than it receives.
I c
(x, y) = Rc
(x, y) · F(x, y)
(7)
where I c
(x, y) is the lightness of the colour channel
c, Rc
(x, y) is the corresponding reflectance, and
F(x, y) is the illumination.
In many cases, the resulting illumination
will be darker than the reflex lightness since most
of the Retinex algorithms are using Gaussian or
Bilateral filters to evaluate the illumination which
means it results in reflecting more light than it
receives i.e., reflectance is more than 1.
In BPF, the illumination is evaluated by
assuming it as the local maxima for each pixel.
Only the neighbours that are brighter than the
central pixel are considered. Illumination can be
evaluated by using the equation below:
Lr(x,y)=
( )
∑ ( ( ( ) ( )) ( ( ) ( ) ( )))( )
(8)
Since already the illumination has been obtained,
the reflectance can be driven by removing
illumination from the reflex lightness:
Rc
(x, y) = I c
(x, y)/Lr (x, y) (9)
The Fig.1 shows an example of image
decomposition through the BPF. From the Fig.1, it
can be said that the reflectance image presents the
details whereas the illumination image presents the
ambience of incident light.
Fig.1. Example for Image Decomposition: (a)
Original image (b)Illumination image
(c)Reflectance image.
III. BI-LOG TRANSFORMATION
Bi-log transformation is made use of here
to perform action mainly on low frequency
information i.e., the negative frequency
components present in the considered image. The
region near zeros are to be highlighted for
enhancement and brightness preservation. Hence,
the region around zeros are enhanced by using this
transformation. The region around zeros are
enhanced by using this transform. The transform
should not suppress the details so that it should be
bright enough and meanwhile the lightness order
should be preserved.
The experimental results show that the log
shape gives good results for several images. The
shape is given by the equation.
Lg(x, y) = log(Lr(x, y) + ε) (10)
where Lr(x,y) is an illumination image, ε is a small
positive constant and is empirically set as 1.
In histogram specification, the intensity of
the processed images appears similar even the input
images looks slightly different based on their
intensities. The illumination can be enhanced bright
enough by making use of log shape histogram
specification. According to the gray-level
distribution of input illumination, the difference is
represented by slightly increasing the pixels of
lower gray-level. The log of the image is taken as
the weight of the histogram which performs well.
The weighted histogram, mp(k), was considered.

61
mp(k) =
∑ ∑ ( ) ( ( ) )
∑ ∑ ( )
(11)
δ(x, y) = 1, for x = y (12)
0, else
where δ is an impulsive function. The modified
histogram considers the values of the gray-levels
along with the numbers of pixels into
consideration.
The Cumulative Density Function (CDF)
of the weighted histogram is given from the
definition of CDF:
cL(v) =∑ ( )
∑ ( ) ( ( )) ∑ ∑ ( ) (13)
U(x,y) = 1, for x >= y (14)
0, else.
Similarly, the CDF of the specified histogram, s(z),
is defined as follows:
cf (z) = ∑ ( )/∑ ( ) (15)
s(z) = log(z + ε), z ∈ N[0, 255] (16)
where z is a non-negative integer within [0, 255], ε
is a small positive constant.
As per histogram specification, definition
needs the purpose of BLT where values of z
satisfies.
cf (zv ) = cL(v), for v = 0, 1, 2, . . . , L – 1
(17)
The values of zv is given by
zv = cf −1
[cL(v)], for v = 0, 1, 2, ..., L – 1
(18)
The final enhanced image is given by:
Lb(x, y) = cf −1
[cL(L(x, y))]
for v = 0,1,2, . . . , L −1 (19)
IV. SYNTHESIS OF REFLECTANCE AND
MAPPED ILLUMINATION
Although the drastic changes in
illumination is one of the drawback for illumination
to display details, illumination is compulsory for
preserving the naturalness of the image. The
mapped illumination is considered mainly to
enhance details and preserve naturalness.
The reflectance R(x, y) and mapped
illumination Lb(x, y) are synthesized together to get
the final enhanced image:
Ic
(x, y) = Rc
(x, y) × Lb(x, y) (20)
It is easy to verify the relative order for the
pixels whose reflectance is fixed at 1 which does
not change since the relative order of lightness in
different local areas of the mapped illumination is
same as that of the original illumination.
V. FLOW CHART
Fig.2. Flow chart of the implemented
algorithm.
V. RESULTS & DISCUSSION
The implemented algorithm has been
tested on the dataset of low contrast and high
contrast gray-scale images. In this paper, four
representatives are presented including Room, sea,
building and girl images. These images are
processed by this algorithm using BPF with and
without using Block processing. The results of this
algorithm in comparison with BPDHE [17]
algorithm are presented in the results Fig.3 to 6
simultaneously.
BPDHE effectively preserves the lightness
order of the input image. It is a global histogram
equalization (HE) algorithm, which is
disadvantageous to highlight the details in areas of
low intensity. The algorithm implemented results in
good performance compared to BPDHE algorithm.
This algorithm is implemented with and
without using Block processing. In block
processing, the processing is done by considering
each block of the image separately where as in
Original Image
Reflectance
Bright-pass filter
Illumination
Final Enhanced Image
Illumination
transformation

62
general the overall image is considered. The results
using Block processing achieves better
performance compared to the results without using
block processing. From the results, it can be seen
that this algorithm preserves the naturalness of the
image while enhancing the low contrast details.
Form observing the results of the girl image, it can
be said that there is no improvement for high
contrast images using this algorithm. This
algorithm works for low contrast images only.
(a) (b) (c) (d)
Fig.3. Results for Image Room (a) Original Image (b) BPDHE Image (c) Implemented algorithm without
using Block processing (d) Implemented algorithm with using Block processing.
(a) (b) (c) (d)
Fig.4. Results for Image Sea (a) Original Image (b) BPDHE Image (c) Implemented algorithm without
(a) (b) (c) (d)
Fig.5. Results for Image Building (a) Original Image (b) BPDHE Image (c) Implemented algorithm
without using Block processing (d) Implemented algorithm with using Block processing.

63
(a) (b) (c) (d)
Fig.6. Results for Image Girl (a) Original Image (b) BPDHE Image(c) Implemented algorithm without
CONCLUSION
Image enhancement plays an important
role in image processing applications. In this paper,
the algorithm is carried out using BPF and Bi-log
transformation. This algorithm is implemented with
and without using Block processing. Both the
results are compared with BPDHE algorithm. This
algorithm with Block processing results in better
performance compared with others. This algorithm
preserves the naturalness while enhancing the low
contrast details.
REFERENCES
[1] Andrea Polesel, Giovanni Ramponi, and V.
John Mathews, ―Image Enhancement via Adaptive
Unsharp Masking‖ IEEE Transactions on Image
Processing, vol. 9, no. 3, March 2000.
[2] J. Alex Stark, ―Adaptive Image Contrast
Enhancement Using Generalizations of Histogram
Equalization‖ IEEE transactions on Image
Processing, vol. 9, no. 5, May 2000.
[3] Sos S. Agaian, Karen Panetta, and Artyom M.
Grigoryan, ―Transform-Based Image Enhancement
Algorithms with Performance Measure‖ IEEE
Transactions on Image Processing, vol. 10, no. 3,
March 2001.
[4] ZhiYu Chen, Besma R. Abidi, David L. Page,
and Mongi A. Abidi, ―Gray-Level Grouping
(GLG): An Automatic Method for Optimized
Image Contrast Enhancement—Part I: The Basic
Method‖ IEEE Transactions on Image Processing,
vol. 15, no. 8, August 2006.
[5] ZhiYu Chen, Besma R. Abidi, David L. Page,
and Mongi A. Abidi, ―Gray-Level Grouping
(GLG): An Automatic Method for Optimized
Image Contrast Enhancement—Part II: The
Variations‖ IEEE Transactions on Image
Processing, vol. 15, no. 8, August 2006.
[6] Bo Ra Lim, Rae-Hong Park, and Sunghee Kim,
―High Dynamic Range for Contrast Enhancement‖
IEEE Transactions on Consumer Electronics, vol.
52, no. 4, November 2006 Contributed Paper.
[7] Sos S. Agaian, Blair Silver, and Karen A.
Panetta, ―Transform Coefficient Histogram-Based
Image Enhancement Algorithms Using Contrast
Entropy‖ IEEE Transactions on Image Processing,
vol. 16, no. 3, March 2007.
[8] M. Abdullah-Al-Wadud, Md. Hasanul Kabir,
M. Ali Akber Dewan, and Oksam Chae, ―A
Dynamic Histogram Equalization for Image
Contrast Enhancement‖ IEEE Transactions On
Consumer Electronics, vol. 53, no. 2, May 2007.
[9] Chun-Ming Tsai and Zong-Mu Yeh, ―Contrast
Enhancement by Automatic and Parameter-Free
Piecewise Linear Transformation for Color
Images‖, Contributed Paper, May 2008.
[10] Jinshan Tang, Xiaoming Liu, and Qingling
Sun, ―A Direct Image Contrast Enhancement
Algorithm in the Wavelet Domain for Screening
Mammograms‖ IEEE Journal of selected topics in
signal processing, vol. 3, no. 1, February 2009.
[11] Jin-Hyuk Hong, Sung-Bae Cho, and Ung-
Keun Cho, ―A Novel Evolutionary Approach to
Image Enhancement Filter Design: Method and
Applications‖ IEEE Transactions on Systems, Man,
and Cybernetics—Part B: Cybernetics, vol. 39, no.
6, December 2009.
[12] Guang Deng, ―A Generalized Unsharp
Masking Algorithm‖ IEEE Transactions on Image
Processing, vol. 20, no. 5, May 2011.
[13] Jae Ho Jang, Sung Deuk Kim, and Jong Beom
Ra, ―Enhancement of Optical Remote Sensing
Images by Subband-Decomposed Multiscale
Retinex with Hybrid Intensity Transfer Function‖
IEEE Geoscience and Remote Sensing letters, vol.
8, no. 5, September 2011.
[14] Shahan C. Nercessian, Karen A. Panetta, and
Sos. S. Agaian, ―Non-Linear Direct Multi-Scale
Image Enhancement Based on the Luminance and
Contrast Masking Characteristics of the Human
Visual System‖ IEEE Transactions on Image
Processing, vol. 22, no. 9, September 2013.
[15] Shuhang Wang, Jin Zheng, Hai-Miao Hu, and
Bo Li, ―Naturalness Preserved Enhancement
Algorithm for Non-Uniform Illumination Images‖
IEEE Transactions on Image Processing, vol. 22,
no. 9, September 2013.
[16] Mila Nikolova, Senior Member, IEEE, and
Gabriele Steidl, ―Fast Hue and Range Preserving
Histogram Specification: Theory and New
Algorithms for Color Image Enhancement‖ IEEE

64
transactions on image processing, vol. 23, no. 9,
September 2014.
[17] H. Ibrahim and N. Kong, ―Brightness
preserving Dynamic Histogram equalization for
Image Contrast Enhancement,‖ IEEE Transactions
On Consumer Electronics, vol. 53, no. 4,
November 2007.
[18] ―Digital image processing,‖ Rafael C
Gonzalez, Richard E. Woods Pearson Education,
2000.

Iisrt zzz bhavyasri vanteddu

More Related Content

What's hot (20)

Viewers also liked (18)

Similar to Iisrt zzz bhavyasri vanteddu (20)

More from IISRT (20)

Recently uploaded (20)

Iisrt zzz bhavyasri vanteddu