Contrast enhancement and brightness preserving of digital mammograms using fuzzy clipped contrast-limited adaptive histogram equalization algorithm

G Model

ARTICLE IN PRESS

ASOC 3443 1–11

Applied Soft Computing xxx (2016) xxx–xxx

Contents lists available at ScienceDirect

Applied Soft Computing journal homepage: www.elsevier.com/locate/asoc

Contrast enhancement and brightness preserving of digital mammograms using fuzzy clipped contrast-limited adaptive histogram equalization algorithm

1

2

3

Q1

4

Sheeba Jenifer a,∗ , S. Parasuraman a , Amudha Kadirvelu b a

5

b

6

School of Engineering, Monash University Malaysia, Bandar Sunway, 46150 Selangor, Malaysia School of Medicine and Health Sciences, Monash University Malaysia, Bandar Sunway, 46150 Selangor, Malaysia

7

8 22

a r t i c l e

i n f o

a b s t r a c t

9 10Q2 11 12 13 14

Article history: Received 19 March 2015 Received in revised form 8 September 2015 Accepted 20 January 2016 Available online xxx

15

21

Keywords: Contrast enhancement Histogram equalization Clip-limit Fuzzy inference system Fuzzy rules

23

1. Introduction

16 17 18 19 20

A novel fuzzy logic and histogram based algorithm called Fuzzy Clipped Contrast-Limited Adaptive Histogram Equalization (FC-CLAHE) algorithm is proposed for enhancing the local contrast of digital mammograms. A digital mammographic image uses a narrow range of gray levels. The contrast of a mammographic image distinguishes its diagnostic features such as masses and micro calcifications from one another with respect to the surrounding breast tissues. Thus, contrast enhancement and brightness preserving of digital mammograms is very important for early detection and further diagnosis of breast cancer. The limitation of existing contrast enhancement and brightness preserving techniques for enhancing digital mammograms is that they limit the amplification of contrast by clipping the histogram at a predefined clip-limit. This clip-limit is crisp and invariant to mammogram data. This causes all the pixels inside the window region of the mammogram to be equally affected. Hence these algorithms are not very suitable for real time diagnosis of breast cancer. In this paper, we propose a fuzzy logic and histogram based clipping algorithm called Fuzzy Clipped Contrast-Limited Adaptive Histogram Equalization (FC-CLAHE) algorithm, which automates the selection of the clip-limit that is relevant to the mammogram and enhances the local contrast of digital mammograms. The fuzzy inference system designed to automate the selection of clip-limit requires a limited number of control parameters. The fuzzy rules are developed to make the clip limit flexible and variant to mammogram data without human intervention. Experiments are conducted using the 322 digital mammograms extracted from MIAS database. The performance of the proposed technique is compared with various histogram equalization methods based on image quality measurement tools such as Contrast Improvement Index (CII), Discrete Entropy (DE), Absolute Mean Brightness Coefficient (AMBC) and Peak Signal-to-Noise Ratio (PSNR). Experimental results show that the proposed FC-CLAHE algorithm produces better results than several state-of-art algorithms. © 2016 Elsevier B.V. All rights reserved.

Q3 24 25 26 27 28 29 30 31

Breast Cancer is the most frequently diagnosed cancer in women [1]. Mammography is proven to be one of the effective diagnostic tools for early detection of breast cancer. A digital mammographic image uses a narrow range of gray levels. The histogram structure of digital mammograms is not well-defined. Both noncancerous and cancerous breast masses appear as white regions in mammographic films. The fatty tissues appear as black regions. The other components of the breast, such as glands, connective

∗ Corresponding author. Tel.: +60 16 2669607. E-mail address: [email protected] (S. Jenifer).

tissue, tumors and calcium deposits appear as shades of gray, more toward the brighter intensity on a digital mammogram [2]. These varied representations of gray levels make digital mammograms as difficult images to interpret. A large number of digital mammograms generated each year need accurate and fast interpretation of images. Non-cancerous lesions can be misinterpreted as a cancer (false-positive value), while cancers may be missed (false-negative value). As a result, radiologists fail to detect 10–30% of breast cancers [3]. Computer aided diagnostic (CAD) systems can help radiologists in accurate interpretation of digital mammograms. The first step in the CAD system for the analysis of digital mammogram images involves pre-processing of the images for contrast enhancement while preserving the brightness of the images.

http://dx.doi.org/10.1016/j.asoc.2016.01.039 1568-4946/© 2016 Elsevier B.V. All rights reserved.

Please cite this article in press as: S. Jenifer, et al., Contrast enhancement and brightness preserving of digital mammograms using fuzzy clipped contrast-limited adaptive histogram equalization algorithm, Appl. Soft Comput. J. (2016), http://dx.doi.org/10.1016/j.asoc.2016.01.039

32 33 34 35 36 37 38 39 40 41 42 43 44 45

G Model ASOC 3443 1–11 2 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111

ARTICLE IN PRESS S. Jenifer et al. / Applied Soft Computing xxx (2016) xxx–xxx

The contrast enhancement and brightness preserving techniques should not deteriorate or destroy the information content in the image for fast and accurate interpretation of digital mammograms. Thus, contrast enhancement and brightness preserving of digital mammograms is very important for early detection and further diagnosis of breast cancer. The fundamental enhancement needed in mammography is an increase in contrast. Contrast between malignant tissue and normal dense tissue may be present on a mammogram, but below the threshold of human perception [4]. Several works have been done in the past for contrast enhancement using histogram equalization of images [5–9]. Generally, histogram equalization stretches the contrast of the high histogram regions, and compresses the contrast of the low histogram regions [10]. As they push the intensities toward the extreme right or the extreme left side of the histogram it causes level saturation effects and when the region of interest occupies a small portion of the image it will not be properly enhanced. As the digital mammograms are textural images, contrast enhancement of these images using the conventional algorithms is very difficult. Adaptive Unsharp Masking (USM) [11] was applied for contrast enhancement. It also lacks in detecting low contrast edges in digital mammograms. Dynamic Histogram Equalization (DHE) [12], is used to eliminate the domination of higher histogram components on lower histogram components in the image histogram and to control the amount of stretching of gray levels for reasonable enhancement of the image features by using local minima separation of histogram. Brightness Preserving Dynamic Histogram Equalization (BPDHE) is an extension method of the DHE and Multi Peak Histogram Equalization with Brightness Preserving (MPHEBP) and this technique divides the input histogram based on local maximum value [13]. BPDHE shown better contrast enhancement compared to MPHEBP and mean brightness preserving compared to DHE. Research [14–30] shows various contrast enhancement methods based on histogram equalization and fuzzy techniques. The research on contrast enhancement of digital mammograms [31,32] using adaptive neighborhood methods are also not immune to noise and produces more artifacts. Contrast improvement technique proposed by Rangayyan et al. [33] improves the contrast of the mammogram image while compromising the naturalness of the original image. Kim et al. [34] used first derivative and local statistics to enhance mammograms. This method could not handle the texture nature of mammogram images. It is more suitable for low degree of gray level discontinuities. Partitioned iterated function systems (PIFS) proposed by Economopoulos et al. [35] is also not suitable for contrast enhancement of digital mammograms because it produces more artifacts. The contrast limited adaptive histogram equalization (CLAHE) algorithm proposed by Zuiderveld et al. [36] has very good results for contrast enhancement of digital mammograms. But it is also not suitable for images of very fine details. Sundaram et al. [37] proposed histogram modified local contrast enhancement (HM-LCE) for mammogram images. This method brings out the local details present in the original image for more relevant interpretation but hidden information is not significantly enhanced. From the literature survey done, it is evident that contrast enhancement without losing relevant information and without artifacts while preserving the naturalness of the original mammogram still remains a challenge. The proposed Fuzzy Clipped Contrast-Limited Adaptive Histogram Equalization (FC-CLAHE) algorithm enhances the local contrast of digital mammograms while preserving the brightness of the mammogram. The contrast is sufficiently enhanced to make the diagnosis more accurate. Clipped Histogram Equalization (CHE) [38–41] methods are used to overcome these problems by restricting the enhancement rate. Clipped Histogram Equalization technique modifies the shape of the histogram of the input images

by minimizing or increasing the value in the histogram’s bins based on a threshold limit before the equalization process. The clipped portion will be redistributed back to the histogram and then histogram equalization is carried out. Clipped Histogram Equalization is far more effective for contrast enhancement than the existing histogram equalization based methods. The major drawbacks of the Clipped Histogram Equalization method are that these methods require manual setting of plateau level of the histogram which is not suitable for automatic systems and some of the methods put weight to the modified histogram. The weight factor also depends on the user. The proposed FC-CLAHE algorithm is based on the observation that the existing histogram equalization techniques limits the amplification of contrast by clipping the histogram at a predefined value called clip-limit. This clip-limit is crisp and invariant to image data. This causes all the pixels inside the window region of the image to be equally affected. Higher values of clip limit result in more contrast and hence this algorithm is not very suitable for real time applications. The proposed algorithm automates the selection of clip-limit which makes it flexible and variant to image data. A fuzzy inference system is designed to automate the selection of clip-limit with a limited number of control parameters. The fuzzy rules are developed to make the clip limit flexible and variant to mammogram data without human intervention. Experiments are conducted using the 322 digital mammograms extracted from MIAS database. Various histogram equalization methods are compared with image quality measurement tools such as Contrast Improvement Index (CII), Discrete Entropy (DE), Absolute Mean Brightness Coefficient (AMBC) and Peak Signal-to-Noise Ratio (PSNR). Subjective evaluation is done with expert radiologists. Experimental results show that the proposed FC-CLAHE algorithm produces better enhanced images than several state-of-art algorithms.

112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144

2. Proposed FC-CLAHE method

145

2.1. Histogram equalization

146

For a given image X =



  X(i, j) with L discrete gray levels 

denoted as X0 , X1 , . . ., XL−1 , the probability density function, p(Xk ) is given by: p(Xk ) =

Nk , N

for k = 0, 1. . ., L − 1

(1)

where Nk represents the number of times the level Xk appears in the input image X and N is the total number of samples in the input image. L is the number of gray levels (L = 256) of the given image. Then, the cumulative density function, c(x) is defined by c(Xk ) =

k 

p(Xj )

for k = 0, 1. . ., L − 1

(2)

147 148 149

150

151 152 153 154

155

j=0

where x = 0, 1,. . ., L − 1. Histogram equalization is a scheme that maps the input image into the entire dynamic range (X0 , XL−1 ) by using the cumulative density function as a transform function. The transformation function f(x) based on the cumulative density function is given as: f (x) = X0 + (XL−1 − X0 ).c(x)

(3)

where (XL−1 ) represents the maximum gray level. The output image produced by histogram equalization is expressed as Y = f (x) =



f (X(i, j))|∀X(i, j) ∈ X



156 157 158 159 160

161

162 163

(4)

164

where (i, j) are the spatial coordinates of the pixel in the image [25].

165

Please cite this article in press as: S. Jenifer, et al., Contrast enhancement and brightness preserving of digital mammograms using fuzzy clipped contrast-limited adaptive histogram equalization algorithm, Appl. Soft Comput. J. (2016), http://dx.doi.org/10.1016/j.asoc.2016.01.039

G Model

ARTICLE IN PRESS

ASOC 3443 1–11

S. Jenifer et al. / Applied Soft Computing xxx (2016) xxx–xxx

h (xk)

h (x

k

)

3

h Clip Limit

Clip Limit

Xk

Xk

X Redistributed Clipped Pixels

0

(a)

L-1

(b)

0

L-1

0

L-1

(c)

Fig. 1. Clipped Histogram Equalization method. (a) Histogram of the original input image. (b) Clipping the histogram based on predefined Clip Limit. (c) Modified histogram after redistribution of the Clipped portion.

166

167 168 169 170 171 172 173 174

175

2.2. Clipped histogram equalization Histogram equalization stretches the contrast of the high histogram regions and compresses the contrast of the low histogram regions. As a result, if the region of interest in an image occupies only a small portion, it will not be properly enhanced during histogram equalization. Clipped histogram equalization methods try to overcome these problems by restricting the enhancement rate [22]. As given in Eq. (3), the enhancement through histogram equalization technique is dependent on c(x). d c(x) = p(x) dx

183

184

Clip Limit =

177 178 179 180 181 182

185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212

Generation of histogram

Calculation of Contrast and Entropy values

(5)

Therefore, the enhancement rate can be limited by restraining the value of p(x) or h(x). Hence the clipped histogram modifies the shape of the input histogram by reducing or increasing the value in the histogram bins based on a threshold limit before the process of equalization. This threshold limit is also known as clipping/clip limit [22]. Fig. 1 shows the steps in Clipped Histogram equalization method. The histogram of the input image is clipped using the clip limit calculated using Eq. (6).

176

Original digital mammogram image

 ϕ    256

+ ˇ. ϕ −

 ϕ  256

Fuzzy Inference System

Selection of fuzzy clipping enhancement parameter

Clipped Histogram

(6)

where ˇ = clipping enhancement parameter [.] denotes truncating the value to the nearest integer ϕ = product of block size The value of 256 denotes the number of bins (0–255). The clipped portion is then redistributed back into the histogram. Histogram equalization is carried out using this modified histogram [25]. The major drawback of the existing clipped histogram equalization method is that the histograms are clipped at a pre-defined clip-limit value. This clip-limit is crisp and invariant to image data. This causes all the pixels inside the window region of the image to be equally affected. Higher values of clip limit result in more contrast enhancement. In most of the cases the user needs to manually set the clip limit which makes these methods not suitable for automatic systems. Self-adaptive plateau histogram equalization (SAPHE) [22] selects the clip limit automatically but the process is complicated and sometimes fails in execution. These algorithms are not suitable for contrast enhancement of digital mammograms as the crisp clip limits equally affect the white and black regions of the image data. Fig. 2 shows the flowchart of the proposed FC-CLAHE algorithm. The proposed FC-CLAHE enhancement method automates the selection of the clip-limit which is relevant and variant to the image data. This makes the algorithm flexible and can be used for real time applications. The histogram of the original image is generated. The values of contrast and entropy of the image are extracted and given as an input to the Fuzzy Interface System. The fuzzy-rules are used to select the fuzzy clipping enhancement parameter. The

Redistribution of clipped pixels

Adjust Fuzzy Rules and Membership functions

No Quality Check Yes Fuzzy Clipped Contrast Enhanced Image Fig. 2. Flow Chart of the proposed FC-CLAHE. Contrast and Entropy values of the digital mammograms are given as an input to the Fuzzy Inference System. Fuzzy rules are developed for the selection of fuzzy clipping enhancement parameter. The clipped pixels in the resultant histogram are redistributed to obtain the fuzzy clipped contrast enhanced image.

new clipped histogram is obtained using the fuzzy clip limit. The quality of the enhanced output image is checked and the fuzzy rules are altered for obtaining the desired results. 2.3. Fuzzy clip based fuzzy inference system The contrast of a mammographic image is the amount to which different diagnostic features such as masses and micro calcifications can be visually distinguished from one another with respect to the surrounding breast tissues. Calcifications in mammograms are

Please cite this article in press as: S. Jenifer, et al., Contrast enhancement and brightness preserving of digital mammograms using fuzzy clipped contrast-limited adaptive histogram equalization algorithm, Appl. Soft Comput. J. (2016), http://dx.doi.org/10.1016/j.asoc.2016.01.039

213 214 215

216

217 218 219 220

G Model

ARTICLE IN PRESS

ASOC 3443 1–11

S. Jenifer et al. / Applied Soft Computing xxx (2016) xxx–xxx

4 Table 1 Linguistic variable and their ranges.

Table 2 Set of derived fuzzy rules for automatic selection of fuzzy clipping enhancement parameter.

Linguistic variable: contrast, C Linguistic value

Notation

Numerical range

Low Medium High

v1 v2 v3

[0,0.3] [0.1,0.6] [0.4,1]

Rule 1: If C is Low and E is Small, then fˇ is Clip Limit High Rule 2: If C is Medium and E is Small, then fˇ is Clip Limit High Rule 3: If C is Low and E is Big, then fˇ is Clip Limit High Rule 4: If C is High and E is Small, then fˇ is Clip Limit Low Rule 5: If C is Medium and E is High, then fˇ is Clip Limit Low Rule 6: If C is High and E is High, then fˇ is Clip Limit Low

Linguistic variable: discrete entropy, E Linguistic value

Notation

Numerical range

Small Big

w1 w2

[0,0.4] [0.2,0.8]

Linguistic variable: fuzzy clipping enhancement parameter (clip limit), fˇ Linguistic value

Notation

Numerical range

Clip Limit Low Clip Limit High

CL1 CL2

[0,0.05] [0.03,0.1]

225

relatively bright regions of calcium as compared with normal breast tissue. Calcifications present within dense masses could present low contrast with respect to their local background [41]. Entropy is the measure the richness of information in a mammographic image after contrast enhancement.

226

N−1  2 i − j P(i, j) Contrast (C) =

221 222 223 224

(7)

i,j=0

227

Entropy (E) =

N−1 

P(i, j)(− ln P(i, j))

(8)

i,j=0 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260

where P(i, j) represents probability of a possible outcome. The range of summation (i, j = 0) to (N − 1) means simply that each cell in the matrix should be considered. Contrast and Entropy are two important properties of the digital image of a mammogram to detect abnormalities. Fuzzy inference system (FIS) considers two input measures: contrast (C) and discrete entropy (E) to achieve fuzzy clipping enhancement parameter (fˇ), which is the output measure of the FIS. In the proposed FIS inputs, x1 ∈ C (Contrast), y1 ∈ E (Entropy) and one output z ∈ fˇ. Tuning of fˇ is based on the fuzzy composition of the C and E. The fuzzy sets of C are Low (v1 ), Medium (v2 ) and High (v3 ) and E are Small (w1 ) and Big (w2 ). The values for v1 , v2 , v3 , w1 and w2 have been fixed using the benchmark dataset with ground truth. The rules constructed using triangular membership function for classifying the fuzzy clipping enhancement parameter into Clip Limit Low (CL1) and Clip Limit High (CL2). Table 1 show the linguistic variables used as inputs and output, corresponding notations and the dynamic numerical ranges for each variable by expert judgment. Six rules were generated to compute the fuzzy clipping enhancement parameter. Eq. (6) shows that the clip limit which depends on two key parameters such as block size and clipping enhancement parameter which are used to control image quality. When a user determines inappropriate parameters, the results of the clipping enhancement algorithms would be worse than that of histogram equalization algorithms. Experimental investigations have been carried out relating the image quality with the block size and the clipping enhancement parameter, which reveals the image quality mainly depends on the clip limit rather than block size. Table 2 illustrate the derived simple yet convincing inference rules employed in the experiment to select the fuzzy clipping enhancement parameter. The membership degree of all rule consequents previously clipped and scaled are combined into a single fuzzy set. Fig. 3

shows an example of FC-CLAHE inference to show how the output of each rule is aggregated into a single fuzzy set for the overall fuzzy output. The result of the defuzzified crisp output z1 is 0.023. It means the fuzzy clipping enhancement parameter (fˇ) is 0.023 for the given values of contrast (0.4) and entropy (0.8). The fuzzy clipping enhancement parameter (fˇ) varies between 0 and 1. Various levels of contrast enhancement an be achieved by varying fˇ. The FC-CLAHE algorithm uses Eq. (9) to calculate the new fuzzy clip limit which is a variant depending on the contrast and entropy of the input image. Fuzzy Clip Limit =

 ϕ   256



+ fˇ. ϕ −

 ϕ 

(9)

256

where fˇ = fuzzy clipping enhancement parameter (ranges from 0 to 1) [.] denotes truncating the value to the nearest integer ϕ = number of pixels per tiles The value of 256 denotes the number of bins. The standard procedure of histogram equalization is to remap grayscales of input image so that the histogram of output image approximates that of the uniform distribution, resulting in the improvement of subjective quality for the output image. For a modified clipped image Xmod k , the probability density function, is given by: Pmod (Xmod

Nmod k , k) = N

for mod k = 0, 1. . .L − 1

(10)

where Nmod k represents the number of times the level Xmod k appears in the input image X and N is the total number of samples in the input image. L is the number of gray levels (L = 255). The transformation function f(x) based on the cumulative density function shown in Eq. (2) is given as: f (x) = X0 + (XL−1 − X0 ).c(x)

(11)

where (XL−1 ) represents the maximum gray level. The output image produced by fuzzy clipped histogram equalization is expressed as Ymod = f (x) =



f (Xmod (i, j))|∀ Xmod (i, j) ∈ Xmod



(12)

261 262 263 264 265 266 267 268 269 270

271

272 273 274 275 276 277 278 279 280 281 282

283

284 285 286 287 288 289 290 291 292 293

where (i, j) are the spatial coordinates of the pixel in the image.

294

3. Performance measures

295

The performance of the proposed FC-CLAHE technique is measured and compared with various histogram equalization methods based on image quality measurement tools such as Contrast Improvement Index (CII), Discrete Entropy (DE), Absolute Mean Brightness Coefficient (AMBC) and Peak Signal-to-Noise Ratio (PSNR). In order to evaluate the competitiveness of the proposed FC-CLAHE method against existing contrast enhancement techniques, the Contrast Improvement Index (CII) is used. Contrast Improvement Index is defined as: CII =

CProposed COriginal

(13)

CProposed and COriginal are the average values of the local contrast in the output and original images, respectively. Discrete Entropy (DE)

Please cite this article in press as: S. Jenifer, et al., Contrast enhancement and brightness preserving of digital mammograms using fuzzy clipped contrast-limited adaptive histogram equalization algorithm, Appl. Soft Comput. J. (2016), http://dx.doi.org/10.1016/j.asoc.2016.01.039

296 297 298 299 300 301 302 303 304

305

306 307

G Model ASOC 3443 1–11

ARTICLE IN PRESS S. Jenifer et al. / Applied Soft Computing xxx (2016) xxx–xxx

5

Fig. 3. The structure of fuzzy inference. (a) Rule 1: If C is Low and E is Small, then fˇ is Clip Limit High. (b) Rule 2: If C is Medium and E is Small, then fˇ is Clip Limit High. (c) Rule 3: If C is Low and E is Big, then fˇ is Clip Limit High. (d) Rule 4: If C is High and E is Small, then fˇ is Clip Limit Low. (e) Rule 5: If C is Medium and E is High, then fˇ is Clip Limit Low. (f) Rule 6: If C is High and E is High, then fˇ is Clip Limit Low.

Please cite this article in press as: S. Jenifer, et al., Contrast enhancement and brightness preserving of digital mammograms using fuzzy clipped contrast-limited adaptive histogram equalization algorithm, Appl. Soft Comput. J. (2016), http://dx.doi.org/10.1016/j.asoc.2016.01.039

G Model ASOC 3443 1–11 6

ARTICLE IN PRESS S. Jenifer et al. / Applied Soft Computing xxx (2016) xxx–xxx

Fig. 4. Comparison of various enhancement techniques using Dense Benign type digital mammogram (mdb099): (a) Original low contrast image. (b) Result of enhancement using Unsharp Masking (USM) technique. (c) Result of enhancement using Histogram equalization (HE) technique. (d) Histogram of the original image. (e) Histogram of the image enhanced using USM technique. (f) Histogram of the image enhanced using HE technique. (g) Result of enhancement using Brightness Preserving Dynamic Histogram Equalization (BPDHE) technique. (h) Result of enhancement using Contrast-Limited Adaptive Histogram Equalization (CLAHE) technique with Clip Limit = 0.01. (i) Result of enhancement using proposed Fuzzy-Clipped Contrast Limited Adaptive Histogram Equalization (FC-CLAHE) technique with Fuzzy Clip Limit = 0.0750. (j) Histogram of the image enhanced using BPDFHE technique. (k) Histogram of the image enhanced using CLAHE technique. (l) Histogram of the image enhanced using the proposed FC-CLAHE technique. The horizontal axis of each histogram plot corresponds to intensity values and the vertical axis corresponds to values of the probability of occurrence of intensity levels.

Please cite this article in press as: S. Jenifer, et al., Contrast enhancement and brightness preserving of digital mammograms using fuzzy clipped contrast-limited adaptive histogram equalization algorithm, Appl. Soft Comput. J. (2016), http://dx.doi.org/10.1016/j.asoc.2016.01.039

G Model

ARTICLE IN PRESS

ASOC 3443 1–11

7

inf inf inf inf inf inf inf inf inf inf inf inf inf inf inf

90.19 88.71 99.64 93.34 89.07 100.70 89.87 90.99 88.82 106.96 89.57 100.62 90.39 87.72 96.67

95.42 94.40 97.91 96.44 102.27 101.42 101.02 96.35 97.03 102.17 104.36 99.78 106.32 96.95 93.07

92.94 91.47 91.99 93.21 98.85 97.09 97.98 89.81 90.53 94.98 99.27 93.26 96.91 92.28 85.92

48.12 47.71 48.00 46.59 50.04 52.75 46.86 47.53 52.14 49.15 49.99 52.62 49.25 49.86 51.78

5.66 6.60 9.22 6.42 5.25 3.52 7.67 6.61 6.54 5.14 5.75 6.25 7.60 8.32 6.95

23.19 22.20 30.30 25.45 21.32 30.66 22.60 23.84 21.90 36.85 21.76 30.15 23.47 21.08 26.06

21.84 23.08 19.74 22.78 20.39 23.55 21.52 23.15 23.22 24.52 23.34 25.75 22.25 21.41 22.18

18.58 18.26 15.95 17.17 17.12 18.73 15.76 17.34 17.36 17.90 17.22 17.60 16.78 16.04 15.26

is used to measure the richness of information in an image after enhancement. DE is defined in Eq. (14).

136.24 132.66 133.07 134.79 138.62 141.14 133.23 135.29 137.32 139.50 137.06 137.79 131.79 134.60 139.32

USM USM

AMBE

AHE

BPDHE

CLAHE

FC-CLAHE

PSNR

AHE

BPDHE

CLAHE

FC-CLAHE

S. Jenifer et al. / Applied Soft Computing xxx (2016) xxx–xxx



4.57 4.60 5.63 5.57 4.85 3.57 5.48 5.23 5.27 3.98 5.32 5.49 5.30 5.58 5.61 4.39 4.47 5.49 5.09 4.58 3.28 5.29 4.80 4.74 3.87 4.68 4.80 4.88 5.44 5.13 3.60 3.91 4.89 4.01 3.61 2.83 4.60 4.13 3.98 3.51 3.82 3.82 4.30 4.81 4.23 2.78 3.05 3.96 3.19 2.82 1.92 3.82 3.24 3.19 2.74 3.17 3.01 3.61 3.93 3.28 3.92 4.23 5.18 4.20 4.02 2.97 5.03 4.41 4.31 3.75 4.17 3.94 4.66 5.24 4.43 2.51 3.40 5.50 4.30 3.46 2.14 3.27 2.78 3.10 2.53 3.19 5.31 4.70 2.86 3.92 2.04 2.35 2.66 2.90 2.52 1.54 1.87 1.73 1.69 1.60 1.95 3.02 2.02 2.01 2.12 0.83 1.21 1.32 1.74 1.14 0.97 1.45 1.00 1.52 0.92 1.58 2.26 1.34 1.54 1.52 0.55 0.70 1.03 0.68 0.81 0.29 1.25 0.40 0.39 0.57 0.92 1.66 1.12 0.88 0.45 1.19 1.24 1.22 1.21 1.21 1.20 1.22 1.21 1.20 1.17 1.21 1.18 1.26 1.20 1.16 mdb099 mdb058 mdb005 mdb075 mdb001 mdb072 mdb209 mdb211 mdb212 mdb213 mdb225 mdb270 mdb300 mdb315 mdb322

FC-CLAHE CLAHE BPDHE AHE USM

DE

FC-CLAHE CLAHE BPDHE AHE USM

CII

Performance measures Image

Table 3

309

N−1

Entropy (E) =

P(i, j)(− ln P(i, j))

(14)

310

i,j=0

where P(i, j) represents probability of a possible outcome. The range of summation (i, j = 0) to (N − 1) means simply that each cell in the matrix should be considered. When the entropy value of an enhanced image is closer to that of original input image, then the details of the input image are said to be preserved in the output image. Absolute mean brightness coefficient (AMBC) is the absolute difference between the mean values of input image X and output image Y to define the normalized absolute mean brightness error (AMBCN ∈ [0,1]). AMBCN (X, Y ) =



1



1 + MB(X) − MB(Y )

(15)

where MB(X) and MB(Y) are the average values of X and Y, respectively. The higher the value of AMBCN , the better is the brightness preservation and vice versa. Peak Signal to Noise Ratio (PSNR) is used to evaluate the degree of similarity between an enhanced image and its original image. PSNR = 10log10

R2 MSE

(16)

where R is the maximum fluctuation in the input image and Mean Square Error (MSE), represents the difference between the enhanced image and its original image. 4. Experiments results and discussions

Q5 Comparative results of different enhancement algorithms using sample images chosen from MIAS database.

308

MIAS database is used for the development and evaluation of the proposed algorithm. The case sample consists of 322 different types of mammographic images with size 1024 × 1024 pixels with 8 bits per pixel. The performance of the proposed fuzzy-based enhancement algorithm FC-CLAHE has been tested on 322 digital mammograms taken from MIAS database. In order to prove the improved performance of the proposed method over conventional and advanced methods, various quantitative performance measures, discussed in the previous section have been used. Fig. 4 shows a sample digital mammogram of the breast cancer type Dense Benign (mdb099.pgm) taken from MIAS database. The original image before enhancement along with the enhanced images using different enhancement techniques and their corresponding histograms are shown. The horizontal axis of each histogram plot corresponds to intensity values and the vertical axis corresponds to values of the probability of occurrence of intensity levels. The proposed method is efficient in terms of visual quality. Fig. 4(a)–(l) shows the enhanced images after applying unsharp masking, adaptive histogram equalization, brightness preserving dynamic fuzzy histogram, contrast limited adaptive histogram equalization and proposed fuzzy clipped-contrast limited adaptive histogram equalization respectively. Visual Quality is more important for contrast enhancement of an image. Fig. 4(a) shows the low contrast original digital mammogram images extracted from MIAS database and Fig. 4(d) shows the histogram corresponding to it. Fig. 4(b) is the equalized image using unsharp masking technique and Fig. 4(e) shows the histogram corresponding to it. Images are slightly enhanced but not much variation in the visual quality. The histograms are much similar without major variations. Fig. 4(c) is the equalized images using adaptive histogram equalization technique and Fig. 4(f) is the histogram corresponding to it. This technique substantially changes the original image but unfortunately over-saturates several areas

Please cite this article in press as: S. Jenifer, et al., Contrast enhancement and brightness preserving of digital mammograms using fuzzy clipped contrast-limited adaptive histogram equalization algorithm, Appl. Soft Comput. J. (2016), http://dx.doi.org/10.1016/j.asoc.2016.01.039

311 312 313 314 315 316 317 318 319 320

321

322 323 324 325 326

327

328 329 330

331

332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364

G Model

ARTICLE IN PRESS

ASOC 3443 1–11

S. Jenifer et al. / Applied Soft Computing xxx (2016) xxx–xxx

8

CII_USM

CII_AHE

CII_BPDHE

CII_CLAHE

CII_FCCLAHE

12 10 8 6 4 2

1 6 11 16 21 26 31 36 41 46 51 56 61 66 71 76 81 86 91 96 101 106 111 116 121 126 131 136 141 146 151 156 161 166 171 176 181 186 191 196 201 206 211 216 221 226 231 236 241 246 251 256 261 266 271 276 281 286 291 296 301 306 311 316 321

0

Fig. 5. Comparison of CII values for all considered contrast enhancement methods. The CII values show that the performance of the proposed FC-CLAHE method is higher when compared to all the other methods. The greater value of CII indicates that the given image quality is better.

DE_USM

DE_AHE

DE_BPDHE

DE_CLAHE

DE_FCCLAHE

10 8 6 4 2 1 6 11 16 21 26 31 36 41 46 51 56 61 66 71 76 81 86 91 96 101 106 111 116 121 126 131 136 141 146 151 156 161 166 171 176 181 186 191 196 201 206 211 216 221 226 231 236 241 246 251 256 261 266 271 276 281 286 291 296 301 306 311 316 321

0

Fig. 6. Comparison of DE values for all considered contrast enhancement methods. The performance of the proposed FC-CLAHE method is higher than all the other methods. This proves the increase in average information and closeness to the original image.

367 368 369 370 371

of the image. The cause of the washed-out appearance is that the middle of the gray scale on the histogram of the equalized image is simply empty. And that is the reason why the visual quality of the equalized images is poor and looks over-enhanced. Fig. 4(g) is the equalized image using BPDHE. The result is poor when compared to other algorithms. Fig. 4(j) is the histogram for the enhanced image using BPDHE technique. Fig. 4(h) is the equalized image using

AMBC_USM

AMBC_AHE

contrast limited adaptive histogram equalization technique and Fig. 4(k) is the histogram corresponding to it. This technique substantially enhances the original image. The visual quality is better when compared to the previous techniques. Fig. 4(i) is the equalized image using the proposed Fuzzy clipped-contrast limited adaptive histogram equalization technique and the histogram corresponding to it. This is more effective than all the previous techniques.

AMBC_BPDHE

AMBC_CLA HE

AMBC_FCCLAHE

200 180 160 140 120 100 80 60 40 20 0

1 6 11 16 21 26 31 36 41 46 51 56 61 66 71 76 81 86 91 96 101 106 111 116 121 126 131 136 141 146 151 156 161 166 171 176 181 186 191 196 201 206 211 216 221 226 231 236 241 246 251 256 261 266 271 276 281 286 291 296 301 306 311 316 321

366

Fig. 7. Comparison of AMBE values for all considered contrast enhancement methods. A median value of AMBE refers to better brightness preservation. Both, very high value and very low value of AMBE represents poor performance in terms of contrast enhancement.

PSNR_USM

PSNR_AHE

PSNR_BDPHE

PSNR_CLAHE

PSNR_FCCLAHE

60 50 40 30 20 10 0 1 8 15 22 29 36 43 50 57 64 71 78 85 92 99 106 113 120 127 134 141 148 155 162 169 176 183 190 197 204 211 218 225 232 239 246 253 260 267 274 281 288 295 302 309 316 323

365

Fig. 8. Comparison of PSNR values for all considered contrast enhancement methods. FC-CLAHE produces better Peak Signal-to-Noise Ratio (PSNR) values when compared to the traditional AHE technique.

Please cite this article in press as: S. Jenifer, et al., Contrast enhancement and brightness preserving of digital mammograms using fuzzy clipped contrast-limited adaptive histogram equalization algorithm, Appl. Soft Comput. J. (2016), http://dx.doi.org/10.1016/j.asoc.2016.01.039

372 373 374 375 376 377 378

G Model ASOC 3443 1–11

ARTICLE IN PRESS S. Jenifer et al. / Applied Soft Computing xxx (2016) xxx–xxx Contrast before Enhancement

Contrast Enhancement using the proposed FC_CLAHE technique

Contrast Enhancement using the USM technique

Contrast Enhancement using the AHE technique

Contrast Enhancement using the BPDFH technique

Contrast Enhancement using the CLAHE technique

9

0.6 0.5 0.4 0.3 0.2 0.1

1 6 11 16 21 26 31 36 41 46 51 56 61 66 71 76 81 86 91 96 101 106 111 116 121 126 131 136 141 146 151 156 161 166 171 176 181 186 191 196 201 206 211 216 221 226 231 236 241 246 251 256 261 266 271 276 281 286 291 296 301 306 311 316 321

0

Fig. 9. Plot of Contrast values for 322 MIAS database images before and after enhancement using various techniques. FC-CLAHE technique provides better contrast enhancement when compared to other techniques.

380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416

Fig. 4(l) shows the corresponding histograms. The histogram corresponding to the equalized images spread more over the full range of gray scale so there is no washed-out appearance in the output images. Fig. 4(a)–(l) illustrates the results obtained from applying various enhancement algorithms. It can be seen from the figures that the visual quality of original image has been greatly enhanced in the resultant images. Quantitative performance measures are very important in comparing different image enhancement algorithms. Besides the visual results, Contrast Improvement Index (CII), Discrete Entropy (DE), Absolute Mean Brightness Coefficient (AMBC) and Peak Signalto-Noise Ratio (PSNR) values are used here for the performance analysis. Table 3 shows the comparative results of different enhancement algorithms using sample images chosen from MIAS database.). The CII values show that the performance of the proposed FC-CLAHE method on the image mdb005 (CIImdb005 = 5.5090) is higher when compared to all the other methods. The greater value of CII indicates that the given image quality is better after enhancement than the original image. Comparisons of Discrete Entropy (DE) shows that the performance of the proposed FCCLAHE method on the image mdb005 (DEmdb005 = 5.6331) is higher than all the other methods. This proves the increase in average information and closeness to the original image. A median value of Absolute Mean Brightness Coefficient (AMBC) refers to better brightness preservation. Both, very high value and very low value of AMBE represents poor performance in terms of contrast enhancement. AMBC is very high for USM (AMBCmdb072 = 141.1416) and AHE (AMBCmdb072 = infinity) techniques. FC-CLAHE maintains a very optimum value for AMBC (AMBEmdb072 = 97.094) which shows that the proposed method preserves the naturalness of the original mammogram image. FC-CLAHE produces better Peak Signal-to-Noise Ratio (PSNR) values when compared to the traditional AHE technique. PSNRmdb072 = 3.5227 using AHE technique and PSNRmdb072 = 18.735 using the proposed FCCLAHE technique. PSNR value using FC-CLAHE is lower than other methods is because of the noisy backgrounds in the resultant image which fairly occupies a large percentage of the image area. Contrast before Enhancement

Figs. 5–8 show plot of the performance measure values of CII, DE, AMBC and PSNR for the 322 digital mammograms enhanced using USM, HE, BPDHE, CLAHE and proposed FC-CLAHE. The proposed technique has significantly larger values of CII and Discrete entropy. The quantitative measure of the contrast improvement can be defined by the Contrast Improvement Index (CII). In Fig. 5, the CII values show that the performance of the proposed FC-CLAHE method is higher when compared to all the other methods. The greater value of CII indicates that the given image quality is better. In Fig. 6, comparisons of Discrete Entropy (DE) show that the performance of the proposed FC-CLAHE method is higher than all the other methods. This proves the increase in average information and closeness to the original image. A median value of Absolute Mean Brightness Coefficient (AMBC) refers to better brightness preservation. Both, very high value and very low value of AMBE represents poor performance in terms of contrast enhancement. In Fig. 7, AMBC is very high for USM and AHE techniques. FC-CLAHE maintains a very optimum value for AMBC which shows that the proposed method preserves the naturalness of the original mammogram image. FC-CLAHE produces better Peak Signal-to-Noise Ratio (PSNR) values when compared to the traditional AHE technique. Fig. 8 shows the comparison of PSNR values. PSNR value using FC-CLAHE is lower than other methods is because of the noisy backgrounds in the resultant image which fairly occupies a large percentage of the image area. Figs. 9 and 10 show the overall contrast enhancement of the original image using the different enhancement algorithms. Contrast enhancement is better for the proposed technique when compared to other existing techniques. 5. Subjective evaluation In order to evaluate the performance of the proposed FC-CLAHE method for the enhancement of digital mammograms, a visual examination was performed before and after image enhancement. Each mammogram was evaluated by eleven observers, consisting of radiologists and technologists blinded to the study. Table 4 shows the comparative results of observer evaluation in visual improvement of digital mammographic images after enhancement using

Contrast aer Enhancement using the proposed FC_CLAHE technique

0.6 0.5 0.4 0.3 0.2 0.1 0 1 6 11 16 21 26 31 36 41 46 51 56 61 66 71 76 81 86 91 96 101 106 111 116 121 126 131 136 141 146 151 156 161 166 171 176 181 186 191 196 201 206 211 216 221 226 231 236 241 246 251 256 261 266 271 276 281 286 291 296 301 306 311 316 321

379

Fig. 10. Plot of original contrast values for 322 MIAS database images and contrast values after enhancement using the proposed FC CLAHE technique. FC-CLAHE technique provides better contrast enhancement when compared to other techniques.

Please cite this article in press as: S. Jenifer, et al., Contrast enhancement and brightness preserving of digital mammograms using fuzzy clipped contrast-limited adaptive histogram equalization algorithm, Appl. Soft Comput. J. (2016), http://dx.doi.org/10.1016/j.asoc.2016.01.039

417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444

445

446 447 448 449 450 451 452

G Model

ARTICLE IN PRESS

ASOC 3443 1–11

S. Jenifer et al. / Applied Soft Computing xxx (2016) xxx–xxx

10

Table 4 Comparative results of observer performance in a scale of 1–10 for the visual improvement of digital mammographic images after enhancement using the proposed CAD techniques. 1 = Very Poor, 10 = Excellent.

453 454 455 456 457 458 459 460 461 462 463

464

465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492

Q4 493 494

Reader

Visual improvement after enhancement. Yes/No?

Visual improvement rating (in a scale of 1–10)

Usefulness of the proposed CAD for detection and classification of digital mammograms (in a scale of 1–10)

1 2 3 4 5 6 7 8 9 10 11

Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes

7 9 8 7 9 9 8 7 9 7 8

7 8 8 7 8 7 8 8 9 7 8

Average

Yes

8

7.7

the proposed CAD techniques. Each reader read two independent images of digital mammograms before and after enhancement. The response were recorded in a standardized form. All the readers agreed of clear visual improvement in mammographic images after image enhancement. In a scale of 1–10 (1 = Very Poor and 10 = Excellent) the average rating was 8. The overall average rating of the proposed CAD for detection and classification of digital mammograms was 7.6. It can be summarized that the availability of enhancement algorithms like FC-CLAHE will facilitate the usage of CAD software in every day clinical practice for accurate interpretation of digital mammograms. 6. Conclusion A novel contrast enhancement algorithm FC-CLAHE for detecting abnormalities in digital mammogram is proposed. The algorithm automates the selection of clip-limit in the traditional CLAHE technique using the Fuzzy inference system. From the investigations, it is evident that the contrast of the image has been improved by preserving the general shape of the original image. When histogram equalization is applied to mammogram images, a narrow range of input intensity values are mapped to wide range of output intensity value. When contrast limited adaptive histogram equalization technique is applied, the histogram is clipped beyond a certain limit. FC-CLAHE technique automates the selection of clip limit. The accumulated leftover sum is then redistributed among all the bins. These accumulated leftover sum amounts to a significant value and when redistributed makes average bin large in height. Now a wide range of input values is mapped to wide range of output values. The slope of the mapping function rises steadily over the entire range of input values. At the cost of clipping, an ideal mapping function is achieved. It not only preserves the brightness of the image but also improves the image contrast and entropy without deterioration of information in the original image. The performance of the proposed technique is evaluated for all the 322 digital mammogram images from MIAS database. Experimental results and subjective evaluation show that the proposed enhanced method is effective and provides better enhancement for various mammogram images. It is evident that the proposed method has increased the detectability of micro calcifications present in mammogram images and is suitable for all types of breast images including fatty, fatty-glandular and dense-glandular mammograms. Uncited reference [42].

References [1] ACS, American Cancer Society, 2014, Available: http://www.cancer.org/. [2] National Cancer Institute. Available: http://www.cancer.gov/cancertopics/ factsheet/NCI, 2014. [3] E.P. Liane, Can computer-aided detection be detrimental to mammographic interpretation? Radiology 253 (1) (2009) 17–22. [4] T.L. Ji, M.K. Sudareshan, H. Roehrig, Adaptive image contrast enhancement based on human visual properties, IEEE Trans. Med. Imaging 13 (1996) 573–586. [5] J.A. Stark, W.J. Fitzgerald, An alternative algorithm for adaptive histogram equalization, Graph Models Image Process. 58 (2) (1996) 180–185. [6] R.B. Paranjape, W.M. Morrow, Rangayyan, Adaptive neighbourhood histogram equalization for image enhancement, Graph. Models Image Process. 54 (3) (1992) 259–267. [7] S.M. Pizer, E.O.P. Amburn, J.D. Austin, Adaptive histogram equalization and its variations, Comput. Vis. Graph. Image Process. 39 (1987) 355–368. [8] J.A. Stark, Adaptive image contrast enhancement using generalizations of histogram equalization, IEEE Trans. Image Process. 9 (2000) 889–896. [9] Shih-Chia Huang, Chien-Hui Yeh, Image contrast enhancement for preserving mean brightness without losing image features, Eng. Appl. Artif. Intell. 26 (2013) 1487–1492. [10] N. Sengee, H.K. Choi, Brightness preserving weight clustering histogram equalization, IEEE Trans. Cons. Electron. 54 (3) (2008) 1329–1337. [11] A. Polesel, G. Ramponi, J. Mathews, Image enhancement via adaptive unsharp masking, IEEE Trans. Image Process. 9 (2000) 505–510. [12] M.A.A. Wadud, M.H. Kabir, M.A. Dewan, O. Chae, A dynamic histogram equalization for image contrast enhancement, IEEE Trans. Cons. Electron. 53 (2) (2007) 593–600. [13] H. Ibrahim, N.S. Pik Kong, Brightness preserving dynamic histogram equalization for image contrast enhancement, IEEE Trans. Cons. Electron. 53 (4) (2007) 1752–1758. [14] M. Kim, M.G. Chung, Recursively separated and weighted histogram equalization for brightness preservation and contrast enhancement, IEEE Trans. Cons. Electron. 54 (3) (2008) 1389–1397. [15] S.D. Chen, A.R. Ramli, Minimum mean brightness error bi-histogram equalization in contrast enhancement, IEEE Trans. Cons. Electron. 49 (4) (2003) 1310–1319. [16] S.D. Chen, A.R. Ramli, Contrast enhancement using recursive mean-separate histogram equalization for scalable brightness preservation, IEEE Trans. Cons. Electron. 49 (4) (2003) 1301–1309. [17] S.D. Chen, A.R. Ramli, Preserving brightness in histogram equalization based contrast enhancement techniques, Digit. Signal Process. 14 (2004) 413–428. [18] M. Wadud, M. Kabir, O. Chae, A spatially controlled histogram equalization for image enhancement, in: 23rd International Symposium on Computer and Information Sciences, 2008, pp. 1–6. [19] D. Sheet, H. Garud, A. Suveer, M. Mahadevappa, J. Chatterjee, Brightness preserving dynamic fuzzy histogram equalization, IEEE Trans. Cons. Electron. 56 (4) (2010) 2475–2480. [20] B.J. Wang, S.Q. Liew, Q. Li, H.X. Zhou, A real-time contrast enhancement algorithm for infra-red images based on plateau-histogram, Infrared Phys. Technol. 48 (2006) 77–82. [21] S.P.K. Nicholas, H. Ibrahim, C.H. Ooi, D.C.J. Chieh, Enhancement of microscopic images using modified self-adaptive plateau histogram equalization, in: IEEE-Computer Society, International Conference on Computer Technology and Development, 2009. [22] C.H. Ooi, N.S. Pik Kong, H. Ibrahim, Bi-histogram equalization with a plateau limit for digital image enhancement, IEEE Trans. Cons. Electron. 55 (4) (2009) 2072–2080. [23] C.H. Ooi, N.A. Mat Isa, Adaptive contrast enhancement methods with brightness preserving, IEEE Trans. Cons. Electron. 56 (4) (2010) 2543–2551.

Please cite this article in press as: S. Jenifer, et al., Contrast enhancement and brightness preserving of digital mammograms using fuzzy clipped contrast-limited adaptive histogram equalization algorithm, Appl. Soft Comput. J. (2016), http://dx.doi.org/10.1016/j.asoc.2016.01.039

495

496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554

G Model ASOC 3443 1–11

ARTICLE IN PRESS S. Jenifer et al. / Applied Soft Computing xxx (2016) xxx–xxx

555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579

[24] C.H. Ooi, N. Isa, Quadrants dynamic histogram equalization for contrast enhancement, IEEE Trans. Cons. Electron. 56 (4) (2010) 2552–2559. [25] K. Liang, Y. Ma, Y. Xie, B. Zhou, R. Wang, A new adaptive contrast enhancement algorithm for infrared images based on double plateaus histogram equalization, Infrared Phys. Technol. 55 (2012) 309–315. [26] Zhiguo Gui, Yi Liu, An image sharpening algorithm based on fuzzy logic, Optik 122 (2011) 697–702. [27] Chang Bu Jeong, Kwang Gi Kim, Tae Sung Kim, Seok Ki Kim, Comparison of image enhancement methods for the effective diagnosis in successive whole-body bone scans, J. Digit. Imaging 24 (2011) 424–436. [28] G. Raju, M.S. Nair, A fast and efficient color image enhancement method based on fuzzy-logic and histogram, Int. J. Electron. Commun. 68 (2014) 237–243. [29] Limimg Hu, H.D. Cheng, Ming Zhang, A high performance edge detector based on fuzzy inference rules, Inf. Sci. 177 (2007) 4768–4784. [30] Hexiang Bai, Yong Ge, Jinfeng Wang, Deyu Li, Xiaoying Zheng, A method of extracting rules from spatial data based on rough fuzzy sets, Knowl.-Based Syst. 579 (2014) 28–40. [31] A.P. Dhawan, G. Buelloni, R. Gordon, Enhancement of mammographic features by optimal adaptive neighbourhood image processing, IEEE Trans. Med. Imaging 5 (1986) 8–15. [32] R. Gordon, R.M. Rangayyan, Feature enhancement of film mammograms using fixed and adaptive neighbourhoods, Appl. Opt. 23 (1984) 560–564. [33] R.M. Rangayyan, L. Shen, Y. Shen, Improvement of sensitivity of breast cancer diagnosis with adaptive neighborhood contrast enhancement of mammograms, IEEE Trans. Inf. Technol. Biomed. 1 (1997) 161–170.

11

[34] Y.T. Kim, Contrast enhancement using brightness preserving bi-histogram equation, IEEE Trans. Cons. Electron. 43 (1) (1997) 1–8. [35] T.L. Economopoulos, P.A. Asvestas, G.K. Matsopoulos, Contrast enhancement of images using partitioned iterated function systems, Image Vis. Comput. 28 (2010) 45–54. [36] K. Zuiderveld, Contrast limited adaptive histogram equalization, in: P.S. Heckbert (Ed.), Graphics Gems IV, Academic Press, Cambridge, MA, 1994, pp. 474–485 (Chapter VIII.5). [37] M. Sundaram, K. Ramar, N. Arumugam, G. Prabin, Histogram modified local contrast enhancement for mammogram images, Appl. Soft Comput. 11 (2011) 5809–5816. [38] A. Raju, G.S. Dwarakish, D. Venkat Reddy, A comparitive analysis of histogram equalization based techniques for contrast enhancement and brightness preserving, Int. J. Signal Process. Image Process. Pattern Recognit. 6 (5) (2013) 353–366. [39] S. Yang, J.H. Oh, Y. Park, Contrast enhancement using histogram equalization with bin underflow and bin overflow, in: International Conference on Image Processing, 2003, pp. 881–884. [40] B.J. Wang, S.Q. Liu, Q. Li, H.X. Zhou, A real-time contrast enhancement algorithm for infrared images based on plateau histogram, Infrared Phys. Technol. 48 (2008) 77–82. [41] T. Kim, J. Paik, Adaptive contrast enhancement using gain-controllable clipped histogram equalization, IEEE Trans. Cons. Electron. 54 (4) (2008) 1803–1810. [42] R.M. Rangayyan, F.J. Ayres, L. Desautels, A review of computer-aided diagnosis of breast cancer: towards the detection of subtle signs, J. Frankl. Inst. (2007) 312–348.

Please cite this article in press as: S. Jenifer, et al., Contrast enhancement and brightness preserving of digital mammograms using fuzzy clipped contrast-limited adaptive histogram equalization algorithm, Appl. Soft Comput. J. (2016), http://dx.doi.org/10.1016/j.asoc.2016.01.039

580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605