Contrast Doubling Technique with Noise Removal Using IASGLC pp 31 - 35

Contrast Doubling Technique with Noise Removal Using IASGLC

Joish George1 Rekha Lakshmanan2 Shine V J3

1
Lecturer, KMEA Engineering College, Ernakulam, Kerala, India
2
Assistant Professor, KMEA Engineering College, Ernakulam, Kerala, India
3
Assistant Professor, Rajagiri School of Engineering and Technology, Ernakulam, Kerala, India
[email protected],[email protected] 2,[email protected] 3
Abstract – This paper presents an integrated technique Fuzzy logic using entropy measure. Indirect
which enables automatic contrast enhancement of an methods, on the other hand, improve the contrast
image. The proposed idea is implemented based on the through exploiting the under- utilized regions of the
concepts of WHS (Weighted Histogram Separation) and dynamic range without defining a specific contrast
IASGLC (Integrated Automatic Selective Gray Level term. Indirect methods can further be divided into
Computing). The entire gray level range can be utilized
efficiently by reducing the group density of histogram
several subgroups: i) techniques that decompose an
components. Two quantitative fidelity measures, image into high and low frequency signals for
Tenengrad criterion (TEN) and Average pixel Distance manipulation, e.g., homomorphic filtering ii)
(D), proves the effectiveness of our method across a wide histogram modification techniques such as
variety of images. Histogram Equalization, Bi – Histogram
Equalization, Adaptive Histogram Equalization,
Histogram Specification, Gray Level Grouping,
Index Terms— Contrast enhancement, automatic and iii) transform-based techniques. Out of these
selective gray-level grouping, adaptive histogram three subgroups, the second subgroup received the
adjustment. most attention due to its straightforward and
intuitive implementation qualities. Contrast
enhancement techniques in the second subgroup
I. INTRODUCTION modify the image through some pixel mapping
such that the histogram of the processed image is
Contrast enhancement has an important more spread than that of the original image.
role in image processing applications. Numerous Contrast enhancement techniques in the first and
contrast enhancement techniques exist in literature, third subgroups often use multi scale analysis to
such as gray-level transformation based techniques decompose the image into different bands and
and histogram processing techniques [1-7]. enhance desired global and local frequencies.
Conventional contrast enhancement techniques These techniques are computationally complex but
generally yield satisfactory results if the proper enable global and local contrast enhancement at the
technique is selected for a given application along same time by enhancing the appropriate scales.
with the proper processing parameters. However The disadvantage of histogram
histogram equalization simply cannot redistribute specification is that it requires manual specification
the histogram components uniformly in most cases, of desired histogram. In order to overcome the
because of the continuous nature of the technique. above shortcomings, the components of the desired
It could cause a washed-out effect and may amplify histogram of the optimal contrast enhancement
background noise along with change in brightness should spread over the full range of the grayscale,
of the image. with the bins being away from each other as far as
Contrast enhancement by machine has possible. The objective of gray-level grouping is to
been a challenging research and development redistribute the histogram components uniformly
problem for scientists and engineers. Poor quality and more efficiently in order to avoid under or over
of the used imaging device, lack of expertise of the contrast and also be able to automatically work on
operator, the adverse external conditions at the time images with different characteristics. Gray level
of acquisition etc are some of the reasons of poor grouping(GLG) based automatic techniques
contrast. This may result in the under-utilization of operate on image pixels and result in the generation
the entire dynamic range. Images with washed-out of uniform histogram by avoiding under or over
and unnatural look may not reveal all the details in contrast problems.
the captured scene. By enhancing the contrast,
these problems are eliminated to obtain a more
Selective gray-level grouping(SGLG)
visually-pleasing or informative image or both.
groups the histogram components in different
segments of the grayscale using different criteria
Contrast enhancement techniques can be
and, hence, is able to enhance different parts of the
broadly categorized into two groups: direct
histogram to various extents.
methods and indirect methods. Direct methods
define a contrast measure and try to improve it.eg-

31
MES Journal of Technology and Management

II. HYBRID TECHNIQUE

For each sub-group at level L, perform the
following steps
The hybrid technique consists of five
procedures, including: weighted histogram
Step 1: Identify non zero histogram components.
separation, histogram clustering, histogram
redistribution [8, 9] histogram appending and
Step2: Cluster the non zero histogram components.
integrated automatic selective gray level
computing.
C. Histogram Redistribution
A. Weighted histogram separation
In order to reduce the group density of
Histogram components of a low contrast
histogram components, weighted histogram
image are concentrated in a narrow gray level
separation technique is used. The following steps
region. A uniform distribution of histogram
describe weighted histogram separation technique.
components is done for utilizing the entire gray
level range. Index mapping function is using for
H11 finding the adjusted positions.
L1
tL, j −1
H' L, j(k' j) = {GL, j(i) | kj = Rd(i * ),0 ≤ i < nL, j}
L2 nL, j −1
H21 H22 where Rd() is a rounding function. Assuming that
the separation level is L, the variables tL,j and nL,j,
are respectively are the adjusted dimension and the
L3 total number of nonzero histogram values in HL,j.

H31 H32 H33 H34

4

Fig.1. Three-level histogram separation structure

2 2
Step 1: Calculate the threshold
4

1
1 t
τ = arg min 1 − ∑ H (i ) where s is
0≤t<s 2 mi = 0
0 1 2 3 4 5 6
histogram dimension(for example s=256
corresponds to the gray-level image) and m is the Fig.3. An example describes the adjusted
total number of pixels. histogram using histogram redistribution.
Step 2: Based on the threshold value, τ divide the
input histogram into two sub-histograms. D. Histogram Appending
Step 3: Perform step 1 and 2 for L levels.
Here we have taken L=3. Histogram separation, clustering and
redistribution can be considered as preprocessing
B. Histogram Clustering steps for performing modified automatic selective
In histogram clustering method, nonzero gray level grouping. The uniformly distributed
histogram components in each sub histograms are histogram components in each subgroups are
grouped separately. combined together to get resultant histogram for
applying IASGLC [10].
4 4
E. Integrated ASGLC
2 2 2 2
4
The algorithm of the integrated automatic
1
4

selective gray level computing technique is

1
described as follows.

0 1 2 3 4 5 6 7 8 9 10 0 1 2 3
(a) (b)
Fig.2. An example describes the processes of the
histogram clustering. (a) Original sub histogram,
(b) Clustered histogram.
 Contrast Doubling Technique with Noise Removal Using IASGLC pp 31 - 35

Start HISTOGRAM VALUE AT POSITION MH IS

REPLACED BY THE average of histogram
value of components ranging from LS
to MH-1
Read the image and obtain its
histogram

Calculate the threshold value for FIND THE SMALLEST GRAY-LEVEL

histogram separation GROUP

Divide histogram into Group the smallest bin with smaller of its two
L levels adjacent neighbors and adjust the gray level bins
to create a new set of gray-level bins

Apply clustering to each sub

histogram

No of N No of bin is
bin= 20 reduced by 1
Redistribute the histogram using
index mapping function

Find the size of gray – level segment

Record the gray level intervals ie it’s right
and left limits

Create look up table for gray-level

Find the position of highest amplitude transformation
histogram component MH

Apply the gray-level transformation function from the

Divide the gray-levels into NS segments computer memory and apply it to the original image to
reconstruct the optimal enhanced image

Find out the gray level segment, S in

which MH is lying. Find its left limit, LS End
and right limit, RS

Fig.4. Flowchart of INTEGRATED method

total number of segments (Here NS=20, the number of
STEP 1: Find the position of highest amplitude gray levels that a human eye can discern)[11]. Left and
histogram component, MH since it represents the right limit of each gray level segment is noted.
background noise.
STEP 3: Find out the gray level segment, S in which
STEP 2: Divide the gray level of histogram highest amplitude histogram component, MH is lying.
components into two or more segments. Let NS be the Let its left and right limit of gray level segment be LS

33
MES Journal of Technology and Management

(a) (b) (c) (d)

(i) (j) (k) (l)

Fig.5. (a),(c) Original girl and moon image are given in the first row resp.(b),(d) Histogram of original girl and moon, (i),(k) After applying
Proposed method (j),(l) Histogram after applying Proposed method.

and RS respectively. This is to locate the gray level STEP 5: The updated histogram is used to perform
segment with noisy background. GLG or fast GLG.

STEP 4: Histogram value at position MH is

replaced by the average of histogram value of III. EXPERIMENT RESULTS
components ranging from LS to MH-1. The
objective is to suppress the background noise and The performance of integrated method has
achieve optimum enhancement of the image. been evaluated, using MATLAB software tool.
Usually high amplitude histogram Tenengrad criterion and Average pixel Distance are
components represent noisy background. In order used to compare the results of contrast
to reduce the noise it should be separate from rest enhancement methods. A histogram based
of the histogram components, although some useful approach is used in these methods. Figure 5 shows
data might be lost in this treatment. the results obtained after applying integrated
In ASGLG the histogram components algorithms. It can be seen from the figure that
from LS to MH are assigned to zero. In order to visual quality of original image has been greatly
reduce the loss of information a single bin for enhanced in resultant images. From the histogram
histogram components from LS to MH and is we can understand the entire gray level range is
assigned average of histogram component values utilized efficiently.
from LS to MH.
Contrast Doubling Technique with Noise Removal Using IASGLC pp 31 - 35

IV. PERFORMANCE ANALYSIS [3]Tarik Arici, Salih Dikbas and Yucel Altunbasak,
A Histogram Modification Framework and Its
Quality measures are very important in Application for Image Contrast Enhancement,
comparing images. Tenengrad criterion and IEEE Transactions on Image Processing, Vol. 18,
average pixel distance are the two quality measures No. 9, September 2009.
mainly used in Integrated method. [4] R. C. Gonzalez and R. E.Woods, Digital Image
TABLE I Processing, 2nd ed.Englewood Cliffs, NJ: Prentice-
Quantitative analysis of Integrated technique for Hall, 2002, ISBN: 0-201-18075-8.
noisy low contrast image(moon, Size 321x 213) [5] Y. Wan, Q. Chen, and B.-M. Zhang, “Image
Method Original IASGLC enhancement based on equal area dualistic sub-
TEN 17733 34941 image histogram equalization method,” IEEE Trans
Consumer Electron., vol. 45, no. 1, pp. 68–75, Feb.
Pixel Dist. 23.92 40.47
1999.
[6] D.-C. Chang and W.-R. Wu, “Image contrast
enhancement based on a histogram transformation
TABLE II of local standard deviation,” IEEE Trans. Med.
Quantitative analysis Integrated technique for Imag., vol. 17, no. 4, pp. 518–531, Aug. 1998.
noisy low contrast camera image (girl, Size 432x [7] Y. Jin, L. Fayad, and A. Laine, “Contrast
283) enhancement by multi-scale adaptive histogram
Method Original IASGLC equalization,” in Proc. SPIE, 2001, vol. 4478, pp.
TEN 55193 86641 206–213.
Pixel Dist. 16.47 33.13 [8] Y. C. Zeng, “Automatic Local Contrast
Enhancement Using Adaptive Histogram
From the results obtained, it is found that Adjustment”, ICME, New York, June 2009.
Tenengrad criterion and average pixel distance [9] S.-C. Pei, Y.-C. Zeng and J.-J. Ding, "Color
value for integrated technique is better than images enhancement using weighted histogram
ASGLG technique. separation," ICIP, Atlanta, GA, USA, pp.2889-
2892, 8-11, Oct. 2006.
[10] Rekha Lakshmanan, Madhu S. Nair, M.
V. CONCLUSION Wilscy, Rao Tatavarti, : Automatic contrast
An integrated technique is proposed which enhancement using Selective Gray-Level
is based on the IASGLC and weighted histogram Grouping: International Journal of Signal and
separation . The proposed technique was tested on Imaging Systems Engineering, Volume 3, Number
a variety of noisy low contrast images, and it was 2, 31 August 2010 , pp. 126-135(10).
demonstrated that there was an improvement in [11] J. Blackwell, “On human vision”, Optical
visual quality, Tenengrad measure and the average Society America, v 36, p624-643, 1946.
pixel distance. IASGLC technique can be
implemented for full automation across a wide
variety of images at faster speeds. Based on the
satisfactory results for IASGLC technique for
automatic contrast enhancement, it is clear that
IASGLC technique can be implemented for full
automation across a wide variety of noisy images at
faster speeds.

REFERENCES
[1] ZhiYu Chen, Besma R. Abidi, David L. Page,
Member 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, Aug. 2006
[2] ZhiYu Chen, Besma R. Abidi, David L. Page,
Member and Mongi A. Abidi, Gray-Level
Grouping (GLG): An Automatic Method for
Optimized Image Contrast Enhancement—Part II:
The Variations. Method IEEE Transactions on
image processing, VOL. 15, NO. 8, Aug. 2006.

35