220

IMAGE CONTRAST ENHANCEMENT BY PRESERVING BRIGHTNESS USING GLOBAL AND LOCAL FEATURES Vijay A. Kotkar1 and Sanjay S. Gharde2 1Research Scholar, Department of Computer Engineering, SSBT’s College of Engg. and Technology, Bambhori, Jalgaon, India 2Assistant Professor, Department of Computer Engineering, SSBT’s College of Engg. and Technology, Bambhori, Jalgaon, India vijaykotkar26@gmail.com, sanjaysgharde358@gmail.com Abstract: Image enhancement is one of the most significant techniques in digital image processing. This paper introduces two image enhancement methods, Weighted of Local and Bidirectional Smooth Histogram Stretching (WLBSHS) and Local then Bidirectional Smooth Histogram Stretching (LBSHS). WLBSHS uses local and global enhancement in weighted approach. Main purpose of local enhancement is sharpening edges of objective and exploring local information. For global enhancement, Bidirectional Smooth Histogram Stretching (BSHS) method is used. We divide the histogram in two parts and use forward and backward gamma transform on these parts, with bin interval control mechanism. We develop the hybrid method which is use to enhance the contrast of image with preserving its brightness by making the mixture of global and local enhancement. In this research we considered AMBE, E, PSNR EME, BR parameters for evaluating the enhanced image. This hybrid method is found to be better than AWIE, AGCID and VHA. Keywords: Image contrast enhancement, global and local features, gamma transform, weighted bidirectional histogram stretching histogram, local bidirectional histogram stretching histogram.

INTRODUCTION

Image enhancement process consist of a collection of techniques that seek to improve the visual appearance of an image or to convert the image to a form better suited for analysis by a human or machine. Image enhancement means as the improvement of an image appearance by increasing dominance of some features or by decreasing ambiguity between different regions of the image. The objective of enhancement is to process an image so that the result is more suitable than the original image for a specific application. Image enhancement is one of the most interesting and visually appealing areas of image processing. Image enhancement approaches fall into two broad categories: spatial domain methods and frequency domain methods [1]. The term spatial domain refers to the image plane itself, and approaches in this category are based on direct manipulation of pixels in an image. Frequency domain processing techniques are based on modifying

the Fourier transform of an image. Enhancement techniques based on various combinations of methods from these two categories are not unusual. Contrast enhancement changes the pixels intensity of the input image to make use of most possible bins [2]. Image enhancement is divided into two categories, local enhancement and global enhancement. Adaptive contrast enhancement (ACE) [3] methods are local enhancement methods. ACE algorithms map the gray values of pixels using the calculation obtained from the local histograms. It utilizes unsharp masking techniques to improve image contrast. In unsharp masking, image is separated into two components, the low-frequency unsharp mask obtained by low-pass filtering of the image, and the high-frequency component obtained by subtracting the unsharp mask from the original image itself. The high-frequency component is then amplified and added back to the unsharp mask to form an enhanced image. Local enhancement based algorithms only make level of pixels in the fixed region, but effectively reduce the impact of other regions, and greatly enhance the local details. In this paper our main attention on global enhancement. So we briefly illustrate some global enhancement methods.

This paper is organized as follows. Section 2 describes the related work. Section 3 describes the proposed LBSHS, WLBSHS methods in detail. Section 4 gives experimental details. Section 5 describes result and discussion.

RELATED WORK

Kim [4] proposed a technique, known as Brightness Preserving Bi Histogram Equalization (BBHE). In this method image histogram is divided into two sub histograms, based on average intensity. Both sub histograms steady independently and then merged. The resultant image mean brightness be positioned between input mean and the middle gray level.

Afterward, Dualistic Sub Image Histogram Equalization (DSIHE) has been proposed by Wan et al [5]. This method is same as BBHE except that, histogram is separated based on median value.

Chen and Ramli proposed a scheme called as Recursive Mean Separate Histogram Equalization (RMSHE) [6]. This method is a recursive version of

221

BBHE. This method separates the histogram into two parts based on the mean input brightness. Again these parts further separated based on their respected mean brightness. If there is r iteration then that will generate 2 histogram pieces. At last each piece independently equalized and then merged. Result is output image include mean brightness close to input mean brightness. Ibrahim and Kong suggest a method Brightness Preserving Dynamic Histogram Equalization (BPDHE) [2] for Image Contrast Enhancement. In which enhanced method is achieved by modifying image histogram. This method is actually an extension of MPHEBP [7] and DHE [8]. In BPDHE first Smooth the histogram with Gaussian filter. Then detect the location of local maximums from the smoothed histogram. For each separation new dynamic range is calculated. On the basis of these new dynamic ranges each separation equalized independently. Finally normalize the image brightness. Ibrahim and Kong also suggest a method Color Image Enhancement Using Brightness Preserving Dynamic Histogram Equalization in which BPDHE utilized for color images [9].

Above describe methods [4]-[8] only give attention to mean brightness value. Although image having less mean brightness error, no assurance to produces natural look and high contrast. Except method [2], [9] all methods based on gray level images, not any instruction available to how they utilized for color images.

He et al propose a method Adjustable Weighting Image Contrast Enhancement Algorithm [10] in which weighted average of histogram equalization, exponential transformation and also input image histogram are combined and achieved contrast enhancement. In this method level of the contrast enhancement is adjusted by changing the weighting coefficients. This is a simple and effective method over histogram enhancement and exponential transformation. But there is not any instruction to how should taken weighted coefficient for dissimilar types of image. Taking exponential transform on intensity value implies lacking color due to the risen of intensity.

Chiu et al proposed a method Efficient Contrast Enhancement Using Adaptive Gamma Correction and Cumulative Intensity Distribution [11] in which image enhancement is achieved by correcting gamma transform. Initially author redistributes probability distribution of luminance channel. Then calculate cumulative probability distribution based on modified probability distribution. Finally perform gamma transform based on modified cumulative probability distribution. The main aim of this method to avoid variation phenomena due to fluctuated probability distribution, during contrast enhancement through gamma transform. This method only suitable for dimmed images, not applicable for blur lighting images and washout type images, because it stretches image histogram only in forward direction.

Xu et al proposed a method Color Image Enhancement by Virtual Histogram Approach (VHA) [12] in which benefits of point operation based image enhancement and histogram modeling both are tries to include. In proposed method first of all mark those intensities on luminance channel that play major roll to construct edges. Then take CDF (Cumulative probability Density Function) of these intensities. Afterward, take CDF of user interested intensity interval and CDF of all intensities. Based on these three CDF, drive a virtual histogram. Applying smoothing procedure on virtual histogram in such a method so that, two consecutive bin does not have difference more than some small constant 2 or 3. Lastly apply boundary limiting mechanism to avoiding spikes at the tail end.

VHA work on luminance channel so that it does not maintain color fidelity of image. Edges in image not sharpen as like in point operation based method. Significant artifacts are generated by this method whenever a light source present in the image.

In this paper we propose two image enhancement methods Weighted of Local and Bidirectional Smooth Histogram Stretching (WLBSHS) and Local then Bidirectional Smooth Histogram Stretching (LBSHS). Both methods not only enhancing global contrast also, sharpen edges, local details and, fully maintaining color fidelity of image. Under WLBSHS method we take weighted result of local and global enhancement. While in LBSHS first we perform local enhancement on input image to get locally enhanced image. Then perform global enhancement on this locally enhanced image to get final result. WLBSHS enhancing image in smoothed approach, while LBSHS produce more sharpen edges and local information.

We carry Local Enhancement (LOCE) sub method for enhancing edges of objects and, local details of image. LOCE is a point operation based method in which we make use of sliding window of size 3 by 3. This window sliding on whole image and expand signals based on local standard deviation. We apply Bidirectional Smooth Histogram Stretching (BSHS) for globally enhancing the image. It divides original histogram into two parts based on dividing point. Then left part stretched by applying modify backward gamma transform and right part is stretched by applying modify forward gamma transform. Each stretched histogram part is smoothed independently and finally merged. LOCE and BSHS both are applying on each color channel of image independently.

PROPOSED METHOD

In this section we propose two methods first is Local Feature Enhancement for exploring local feature of image. Second is Bidirectional Smooth Histogram Stretching for global enhancement of image. We merge result of both procedures by two ways. One is combining in weighted manner to get Weighted of local

222

and Bidirectional Smooth Histogram Stretching (WLBSHS). Other one is first perform Local Enhancement on original image to get locally enhanced image. In second step Perform BSHS on locally

enhance image to get Local then Bidirectional Smooth Histogram Stretching (LBSHS). WLBSHS and LBSHS are useful for different scenarios.

Fig. 1 Block diagram of Proposed Method.

Under local enhancement, we calculate the local feature of image. x(i, j) represents a pixel value of image. f(i, j) denote the enhanced value of x(i, j). w denotes the weighting coefficient. g(i, j) and h(i, j) represents output of WLBSHS and LBSHS methods respectively. Local Feature Enhancement

In local feature enhancement, we calculate the local feature of image. Let x(i, j) represents a pixel value of image. The local area is defined as a (2n+1)×(2n+1) window centered at (i, j). Where n is an odd number. The local mean, i.e. the low-frequency component, of a pixel (i, j) can be computed as

m =1

(2n + 1) x(k, l) (1)

And the local variance calculated as σ

=1

(2n + 1) [x(k, l)− m (i, j)]

(2)

Let f(i, j) denote the enhanced value of x(i, j) then.

f(i, j) = x(i, j) + ( x(i, j) − m ) )σ

max(σ ) (3)

Or in compact form f(i, j) = L x(i, j) (4)

Here L(.) denote local transformation. Bidirectional Smooth Histogram Stretching (BSHS) for Global Enhancement

In this procedure, we get global enhancement. Because the G channel is a good estimate of the luminance signal

[9] so we show G channel histogram under different stages of BSHS procedure in Fig. 1 Steps regarding BSHS are as follows: 1. Initially we find a dividing point r in histogram

from which one can separate the histogram into two parts.

2. Left part of histogram P (r ) is from starting point of histogram up to dividing point of histogram.

3. Right part of histogram P (r ) started just after from dividing point to end point of histogram (dividing point is excluded).

4. Calculate biased cumulative density function C (r ) and C (r ) for left and right histograms respectively. These functions are used to construct modify backward gamma transform and, modify forward gamma transform respectively.

5. Stretching Left part of histogram P (r ) we utilize modify backward gamma transform. It stretches histogram towards backward direction i.e. from dividing point r to start bin 0.

6. Stretching Right part of histogram P (r ) we utilize modify forward gamma transform. It stretches histogram towards forward direction i.e. from dividing point r to last end L-1 bin.

7. Apply smoothing procedure separately on each histogram.

8. Merge both histograms to generate modified starched histogram.

9. Remap original image pixels value according to adapted histogram to get globally enhanced image.

Let r є [0, L −1] be the kth intensity level, and then the probability density function (PDF) P(r ) is defined as

P(r ) =nn

(5)

Input Image x(i, j)

Local feature Enhancement

f(i, j)

BSHS y(i, j)

Local feature Enhancement

f(i, j)

BSHS y(i, j)

w

Output of

LBSHS h(i, j)

Output of

WLBSHS g(i, j)

1-w

223

for k = 0,1,….. ,L −1 Where n represents the number of occurrences of r and n the total number of samples in the input image. Here P(r ) is called the histogram of the input image. Based on the theory of probability, the Cumulative Density Function (CDF) is defined as

C(r ) = P r (6)

for k = 0,1,2,…L-1.

Note that C(r ) =1 by definition. Let m denote mean of intensity value of image x(i, j). Where i & j are pixel coordinates and r = 0, r = L − 1. m є {r , r , r ,… r } where r <r < r ,… r < r . And r є {r , r , r , … r } is dividing bin from, which histogram of image i.e. P(r ) is decomposed into two sub histogram left sub histogram P (r ) for k=0,1,2,…d and right sub histogram P (r ) for k=d+1, d+2 …L-1.

(a)

(b) (c) (d)

(e) (f) (g) Fig. 2 G-channel histogram: (a) Original image (b) Modify backward gamma transform on left histogram (c) Modify forward gamma transform on right histogram (d) Combine both histogram after modified gamma transform (e) Smoothed modify backward gamma transform on left histogram (f) Smoothed modify forward gamma transform on right histogram (g) Combine both smoothed left and right histogram.

224

Here dividing bin r is calculated as

r = arg min abs C(r ) −m

L − 1 (7)

By definition P (r ) ∪ P (r ) = P(r ) (8)

P (r ) ∩ P (r ) = ϕ (9)

C (r ) = P (r ) (10)

Where k є {0,1,2….d}

C (r ) = P r (11)

Where k є {d+1,d+2,….L-1} r = min(r ) (12)

Where P(r ) ≠ 0 Here bsb is a bias value proportional to number of

empty bin before histogram. This bias is useful to more stretch the left part of histogram over starting empty bins.

bsb =r

r + (L − 1) − r (13)

C (r ) = (P (r )− bsb) (14)

Where k є {0,1,2….d}

r = max(r ) (15) Where P(r ) ≠ 0 Here bsf is a bias value proportional to number of

empty bin after histogram. This bias is useful to more stretch the right part of histogram over ending empty bins.

bsf =(L− 1) − r

r + (L − 1) − r (16)

C (r ) = (P (r ) + bsf) (17)

Where k є {d+1,d+2,….L-1} T (r ) stretch left sub histogram P (r ) in

backward direction i.e. from dividing bin r to initial bin 0.

T (r ) = rrr

( ( ) ) (18)

T (r ) stretch right sub histogram P (r ) in forward direction i.e. from dividing bin r to last bin L-1.

푇 (푟 ) = (퐿 − 1)푟

(퐿 − 1)

( ( ) ) (19)

푇 (푟 ) = {푟 , 푟 , 푟 , … … . 푟 } (20) Here r < r < r … … . r < r (21)

r ≥ 0 (22) r = r (23)

T (r ) = {r , r , r , … … . r } (24) r < r < r , … … . r < r (25)

r > r (26) r ≤ L − 1 (27)

Function Lt( ) is used for limit the interval between two consecutive bines, so that maximum interval

between two bins cannot exceed more than △ . In our experiment △ =2.

Lt T (r ) = r , r , r , … … . r (28) r < r < r , … … . r < r (29)

r ≥ 0 (30) r = r (31)

r − r ≤△ (32) Where i=0,1,2…..d-1 Lt T (r ) = r , r , r , … … . r (33) r , r , r , … … . r < r (34)

r = r +△ (35) r ≤ L − 1 (36) r − r ≤△ (37)

Where i = (d+1),(d+2),…(L-2)

Lt T (r ) ∩ Lt T (r ) = 휙 (38) Lt T (r ) ∪ Lt T (r ) = T (r ) (39)

Suppose y (i, j) is enhanced image then y(i, j) = T x(i, j) (40)

Weighted Enhancement of Local and BSHS

g(i, j) = w . f(i, j) + (1− w). y(i, j) (41) Where w є [0, 1]

g(i, j) =Mean(x(i, j))Mean(g(i, j))

g(i, j) (42)

Local then BSHS Enhancement

h(i, j) = T L x(i, j) (43)

h(i, j) =Mean(x(i, j))Mean(h(i, j)) h(i, j) (44)

EXPERIMENTAL DETAILS

In this section, five image enhancement algorithms are carried out in our experiments AWICE [10], AGCCID [11] and VHA [12] and proposed methods ‘LBDSH’, ‘WLBDSH’. Subjective and Objective assessments both are done on a group of images. Three groups of images are taken and Fig. 2 to 6 show the results of image enhancement with their green channel histogram for these images. Under Objective assessments five quantitative measurements are utilized to show strength of proposed methods. Table 1 describes all parameters values taken in our experiment. These parameters are AMBE, E, PSNR, EME and BR.

RESULT AND DISCUSSION

Subjective Assessments

As show in Fig. 3 AWICE method increase the contrast but not preserve natural look of image. AGCCID method over exposes the image. VHA increase background contrast of image but still lose detail.

225

LBDSH and WLBDSH both enhancing contrast with preserving natural visibility of image. All local details clearly enhanced in both images. LBDSH enhance local details in little bit highlighted manner and WLBDSH enhance local details in smother manner.

Another simulation results show in Fig. 5 AWICE not preserving natural color combination of image see in middle and boundary area. Although in

first view image show rich details, but actually it wider spots, holes, and lines rather than sharpening. AGCCID over enhances the image and it not visually pleasing. VHA little bit succeed to preserving color of image. LBDSH and WLBDSH both sharpening details with preserving natural look of image. Here LBDSH sharpening details in more contrasted manner.

(a) Original image (b) AWICE (c) AGCCID

(d) VHA (e) LBDSH (f) WLBDSH

Fig. 3 Results of image - Passport Photograph

(a) Original image

(b) ) AWICE

(c) AGCCID

(d) VHA

(e) LBDSH

(f) WLBDSH

Fig. 4 G-channel histogram of image: Passport Photograph

226

(a) Original image (b) AWICE (c) AGCCID

(d) VHA (e) LBDSH (f) WLBDSH

Fig. 5 Results of image - Historical Building

(a) Original image

(b) AWICE

(c) AGCCID

(d) VHA

(e) LBDSH

(f) WLBDSH

Fig. 6 G-channel histogram of image: Historical Building

Preserving the overall shape of histogram,

corresponding to their bins and smoothness of histogram is responsible for natural look of image. And also utilization of more bins by image histogram showing good enhancement of image. By observing Fig. 4 and Fig. 6 AWICE although utilize almost all bins but it not preserving the histogram shape with their

respective bins. It also shows zigzag nature. AGCCID compress histogram at higher bins and not utilizing starting bins. VHA also not utilize full bins except Fig. 6 (d). Due to more zigzag nature some time it over expose small details. In LBDSH and WLBDSH both are utilizing all bins with preserving shape, smoothness and symmetry of histogram in all cases.

227

Objective Assessments Here we done objective quality assessments based on five parameter Absolute Mean Brightness Error (AMBE) [13], [14] Discrete Entropy (E), Peak Signal-to Noise Ratio (PSNR) [14], Measure of Enhancement (EME) and Brenner’s measure (BR).

By observing Table 2 and Table 3 we can see that LBDSH always generate very high value of EME and BR compare to all methods. And lowest AMBE value compares to first three methods AWIE, AGCCID, VHA. It also shows high range value of E and PSNR.

WLBDSH always generate lowest AMBE value compare to all methods. In almost all cases it show highest PSNR value and high range value of E and EME. LBDSH and WLBDSH both enhance the image details, with preserving natural look.

Common assumption for expression (45) to (50) M≔ Number of pixel row in image N≔ Number of pixel column in image MN≔ Total number of pixel in image I (i, j)≔ (i, j)’th pixel value of input image I (i, j)≔ (i, j)th pixel value of output image I(i, j)≔ (i, j)th pixel value of image L-1≔ Maximum possible pixel value

Method/Parameter Weighted Parameter Window Size Maximum bin interval Other Parameter

AWIE қ=3, λ=1 - - ε=0.009 AGCID - - - -

VHA w=2, v=2 3×3 t=2 P =16, P =40, T =3 LBDSH w=0.5 3×3 △ =2 -

WLBDSH w=0.5 3×3 △ =2 -

Table 1: Parameters used for all method.

Method/Criteria AMBE E PSNR EME BR Original 0 7.3137 Inf 35.2654 657.2504 AWIE 34.8053 7.4273 16.5026 41.3977 824.3338

AGCID 37.9591 7.1374 15.9571 35.9077 1014.3 VHA 5.2827 7.2507 24.1889 33.0475 428.5282

LBDSH 0.3139 7.3410 19.5656 84.3178 2256.6 WLBDSH 0.0560 7.5174 24.6698 43.9301 1482.5

Table 2: Performance for image Passport Photograph.

Method/Criteria AMBE E PSNR EME BR

Original 0 7.7808 Inf 93.9068 1023.3 AWIE 9.8127 7.5879 27.0341 52.6873 987.1799

AGCID 34.2365 7.5591 15.7761 95.7148 1731.6 VHA 13.1609 7.6805 18.1787 53.8198 760.4839

LBDSH 0.0145 6.9427 19.0693 192.7546 3695.9 WLBDSH 0.0146 7.5606 24.3729 130.3341 2193.3

Table 3: Performance for image building of Historical Building.

AMBE. It is an absolute mean difference between input and output image brightness. Lower AMBE value indicates that the mean brightness better preserved.

AMBE =1

MNI (i, j) − I (i, j) (45)

E. The discrete entropy is used to measure the content of an image. Generally, the higher value of discrete entropy indicates image is fulfilling with richer details.

E = P(r )log P(r ) (46)

Where r represent color bin and P(r ) is normalize count pixels having color r .

PSNR. The PSNR stands for Peak Signal-to-Noise Ratio. Noise is taken as root mean-squared error (RMSE) Higher PSNR value represents greater image quality.

PSNR = 20log255

RMSE (47)

Where RMSE is, defined as

RMSE =1

MN[I (i, j) − I (i, j)] (48)

EME. This value is an approximation of the averaged contrast in the image. The EME approximates an average contrast by partitioning the image into numbers of non-overlapped sub-blocks of size k by k . And then finding a measure based on ratio of the minimum

228

and maximum gray values in each sub-block. Finally averages them to generate the result. Higher EME value represents local details are well enhanced in image.

EME =1

Mk

Nk

20InMax I ( , ) + conMin I ( , ) + con

(49)

Where con is small value for avoiding zero values. I ( , ) is image block of size k by k and (i, j) represents block index. In our experiment we take con=0.0001 and k =8, k =8. Max I ( , ) and Min I ( , ) return maximum and minimum value for block indexed by (i, j). ⌊. ⌋ Represents floor function. BR. Brenner’s measure is used for sharpness assessment. Higher value of BR indicates more sharpness of image.

BR = I(i, j) − I(i + 2, j) (50)

FUTURE SCOPE

For global enhancement, BSHS method is work on 1-D histogram and it enhancing global features of image. We can extend it to work on a 2-D histogram of the input image constructed using a mutual relationship between each pixel and its neighboring pixels. This mutual relationship may contain statistical information or respective color information of neighbor pixels. So by this way it is possible to enhancing local features as well as global features of image simultaneously. CONCLUSION

The strength of our proposed approach lies in the interpolation of two methods, local enhancement and Global enhancement. Local enhancement achieved through our local standard deviation based formula. Global enhancement achieved by smoothly stretching histogram in both directions using modified gamma transform. Finally merge them weighted approach and one by one approach in WLBDSH and LBDSH respectively. The experimental result based on subjective and objective assessments approve that proposed methods is effective and flexible. The AMBE value of WLBDSH and LBDSH are 0.0560 and 0.3139 respectively, these values are low as compare to other methods. EME and BR values of WLBDSH are 43.9301 and 1482.5 respectively, for LBDSH are 84.3178 and 2256.6 respectively, these values are high as compare to other methods. E and PSNR values of WLBDSH are 7.5174 and 24.6698 respectively, for LBDSH are 7.3410 and 19.5656 respectively, these values are high as compare to other methods. This novel method is use for enhancing the better quality image with dynamic range.

REFERENCES

[1]Rafael C. Gonzalez, and Richard E. Woods (2002): ‘Digital Image Processing’, 2nd edition, Prentice-Hall of India: New Delhi.

[2]H. Ibrahim and N. S. P. Kong (2007): ‘Brightness preserving dynamic histogram equalization for image contrast enhancement’, IEEE Transactions on Consumer Electronics, vol. 53, no. 4, pp. 1752-1758.

[3] J. Alex Stark (2000): ‘Adaptive image contrast enhancement using generalizations of histogram equalization’, IEEE Trans. Imag. Proc., vol.9, No.5, pp. 889-896.

[4]Yu Wan, Qian Chen and Bao-Min Zhang (1999): ‘Image enhancement based on equal area dualistic sub-image histogram equalization method’, IEEE Trans. Consumer Electron., vol. 45, no. 1, pp. 68-75.

[5]Soong-Der Chen, and Abd. Rahman Ramli (2003): ‘Minimum mean brightness error bi-histogram equalization in contrast enhancement’, IEEE Trans. Consumer Electron., vol. 49, no. 4, pp. 1310-1319.

[6]Soong-Der Chen and Abd. Rahman Ramli (2003): ‘Contrast enhancement using recursive mean-separate histogram equalization for scalable brightness preservation’, IEEE Trans. Consumer Electron., vol. 49, no. 4, pp. 1301-1309.

[7]K. Wongsritong, K. Kittayaruasiriwat, F. Cheevasuvit, K. Dejhan and A. Somboonkaew (1998): ’Contrast enhancement using multipeak histogram equalization with brightness preserving’, IEEE Asia-Pacific Conference on Circuit and System, pp. 455-458, 24-27.

[8]M. Abdullah-Al-Wadud, Md. Hasanul Kabir, M. Ali Akber Dewan, and Oksam Chae(2007): ‘A dynamic histogram equalization for image contrast enhancement’, IEEE Trans. Consumer Electron., vol. 53, no. 2, pp. 593-600.

[9]Nicholas Sia Pik Kong and Haidi Ibrahim (2008): ‘Color Image Enhancement Using Brightness Preserving Dynamic Histogram Equalization’, IEEE Transactions on Consumer Electronics, Vol. 54, No. 4, pp.1962-1968.

[10]Renjie He, Sheng Luo, Zhanrong Jing and Yangyu Fan (2011): ‘Adjustable Weighting Image Contrast Enhancement Algorithm and Its Implementation’, IEEE Conference on Industrial Electronics and Applications, pp.1750-1754.

229

[11]Yi-Sheng Chiu, Fan-Chieh Cheng and Shih-Chia Huang (2011): ‘Efficient Contrast Enhancement Using Adaptive Gamma Correction and Cumulative Intensity Distribution’, IEEE International Conference on Systems, Man, and Cybernetics (SMC), pp. 2946 -2950.

[12]Zhengya Xu, Hong Ren Wu and Xinghuo Yu (2010): ‘Colour Image Enhancement by Virtual Histogram Approach’, IEEE Transactions on Consumer Electronics, Vol. 56, No. 2.

[13]M. Kim and M. G. Chung (2008): ‘Recursively separated and weighted histogram equalization for brightness preservation and contrast enhancement’, IEEE Trans. Consumer Electron., vol. 54, no. 3, pp.1389–1397.

[14]Pei-Chen Wu, Fan-Chieh Cheng and Yu-Kumg Chen (2010): ‘A Weighting Mean-Separated Sub-Histogram Equalization for Contrast Enhancement’, International Conference on Biomedical Engineering and Computer Science (ICBECS), pp. 1-4, 23-25.