Scholarly article on topic 'Mammographic Image Enhancement Using Indirect Contrast Enhancement Techniques – A Comparative Study'

Mammographic Image Enhancement Using Indirect Contrast Enhancement Techniques – A Comparative Study Academic research paper on "Materials engineering"

CC BY-NC-ND
0
0
Share paper
Academic journal
Procedia Computer Science
OECD Field of science
Keywords
{"Image enhancement" / "Histogram equalization" / CLAHE / BPHE / RMSHE / MMBHE / "Effective measure of enhancement ;"}

Abstract of research paper on Materials engineering, author of scientific article — K. Akila, L.S. Jayashree, A. Vasuki

Abstract Contrast enhancement is an important issue in the field of mammographic image processing. It can be classified into two categories: direct contrast enhancement and in direct contrast enhancement. Indirect contrast enhancement involves in modifying histogram of the image. Histogram equalization (HE) is the simplest indirect contrast enhancement technique which is widely used for contrast enhancement. Many variants of HE are proposed so far. Comparison of these techniques is significantly essential in deciding a suitable algorithm for enhancement and further processing. In this paper we applied few indirect contrast enhancement techniques namely histogram equalization, CLAHE, BBHE, RMSHE, MMBEBHE to preprocess the mammographic images. The performance of the methods is measured using effective measure of enhancement (EME) and peak signal to noise ratio (PSNR).

Academic research paper on topic "Mammographic Image Enhancement Using Indirect Contrast Enhancement Techniques – A Comparative Study"

CrossMark

Available online at www.sciencedirect.com

ScienceDirect

Procedia Computer Science 47 (2015) 255 - 261

Mammographie image enhancement using indirect contrast enhancement techniques - A comparative study

K.Akilaa, L.S.Jayashreeb, A.Vasukic

a Mechatronics Engineering, Kumaraguru college of Technology, Coimbatore, India b Computer Science Engineering, RVS College of Engineering and Technology, Coimbatore, India c Electronics and Communication Engineering, Kumaraguru college of Technology, Coimbatore, India

Abstract

Contrast enhancement is an important issue in the field of mammographic image processing. It can be classified into two categories: direct contrast enhancement and in direct contrast enhancement. Indirect contrast enhancement involves in modifying histogram of the image. Histogram equalization (HE) is the simplest indirect contrast enhancement technique which is widely used for contrast enhancement. Many variants of HE are proposed so far. Comparison of these techniques is significantly essential in deciding a suitable algorithm for enhancement and further processing. In this paper we applied few indirect contrast enhancement techniques namely histogram equalization, CLAHE, BBHE, RMSHE, MMBEBHE to preprocess the mammographic images. The performance of the methods is measured using effective measure of enhancement (EME) and peak signal to noise ratio (PSNR).

© 2015 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license (http://creativecommons.Org/licenses/by-nc-nd/4.0/).

Peer-reviewunder responsibilityoforganizing committee ofthe Graph Algorithms,HighPerformanceImplementations and Applications(ICGHIA2014) _

Keywords: Image enhancement; Histogram equalization; CLAHE; BPHE; RMSHE; MMBHE; Effective measure of enhancement;

1 Introduction

Breast cancer is the second leading disease causing death in women, next to lung cancer [24]. Detection and diagnosis of breast cancer in its early stage increases the chances for successful treatment and complete recovery of the patient. Screening mammography has been considered as the reliable imaging system for earlier detection of breast cancer [1]. The subtle signs of cancer such as masses, calcifications are difficult to be detected by the radiologist because mammograms are low contrast [2] which is depicted from Fig. 1. Contrast between malignant tissue and normal dense tissue may be present on a mammogram but below the threshold of human perception.

1877-0509 © 2015 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).

Peer-review under responsibility of organizing committee of the Graph Algorithms, High Performance Implementations and Applications (ICGHIA2014) doi: 10.1016/j.procs.2015.03.205

b 2500

Fig.1. (a) Original mammographic image and (b) its histogram

Hence, the fundamental enhancement needed in mammography is an increase in contrast to enhance image feature against its background to visualize the image properties in an open eye. Various contrast enhancement techniques has been proposed so far [2 - 6, 9 -16].

Direct contrast enhancement establishes a criterion of contrast measure and enhances the images by improving the contrast directly. Establishment of a suitable image contrast measure is the key step of direct image enhancement.Cheng et al proposed Adaptive fuzzy logic contrast enhancement method which is based on fuzzy entropy principle which transforms the image to a fuzzy domain, and computes the fuzzy entropy and measures the local contrast [3].Rangayyan et al proposed Adaptive neighborhood contrast enhancement technique on measurement of local contrast [4]. In [5], Tang et al, proposed a multiscale local contrast measure in the wavelet domain which enhances the details in different scales. This makes the method suitable for the detection of calcifications that exist in different scales. Yicong Zhou et al suggested HVS based contrast enhancement which separates the abnormal regions without using any thresholding or segmentation algorithm. This feature is useful for automatic detection of breast cancer in the CAD systems [6].

3 Indirect contrast enhancement

Indirect contrast enhancement cannot manipulate the image contrast directly; rather it modifies the histogram of the image and thus increases contrast. Histogram equalization techniques are the popular indirect contrast enhancement methods.

In this paper the general Histogram Equalization (HE), Contrast Limited Adaptive Histogram Equalization(CLAHE), Brightness Preserving Bi-Histogram Equalization (BBHE), Minimum Mean Brightness Error Bi-Histogram Equalization (MMBEBHE), Recursive Mean Separate Histogram Equalization(RMSHE) are analyzed.

3.1 Histogram Equalization (HE)

Histogram Equalization [7, 8, 17-20] is a posteriori modelling technique which maps the input gray levels to a gray level proportional to its cumulative density and hence, the probability of each gray level in resulting image is uniformly distributed. The output image histogram should ideally contain an equal number of pixels at every discrete gray level value. The histogram equalization method is global and blind such that it does not take input visual detail in to account while enhancing the image. It results in excessive contrast enhancement, which causes the unnatural look and visual artifacts of the processed image [7].

3.2 Contrast Limited Adaptive Histogram Equalization (CLAHE)

CLAHE [7-9, 19-21] is a variant of adaptive histogram equalization. It divides the original image into several non overlapping sub images. Histogram of the sub images are clipped to limit the amount of enhancement of each pixel and then equalized. The details in the image appear clearly relative to the

2 Direct Contrast Enhancement Techniques

background [8]. The same time the background of image is enhanced equally as the foreground of the image which leads to the high contrast output image [9].

3.3 Brightness Preserving Bi-Histogram Equalization (BBHE)

BBHE [13 -15] divides the original gray level image into two sub-levels based on the mean brightness of the image. One of the sub level image have range from minimum gray level to mean and the other ranges from mean to maximum. After this separation process, this technique equalizes histogram of each sub-level images independently which results in brightness preserved contrast enhanced image.

Let Xm denote the mean of the image X and assume that Xm C {X0, X1..... XL-1}. Based on the mean Xm

the input image is divided into two sub level images XL and Xy.The transform functions [13] for the sub images are defined as

fL(X) = X0 + (-X0)CL(X) fu(X) = Xm+1 + (XL_, -Xm+l)Cu(X)

Where CL(X) and CV(X) is the respective cumulative density functions for XL and Xu. The output image(Y) of BBHE, is expressed as

Y = fL( Xl ) U fu (Xu ) 3.4 Minimum Mean Brightness Error Bi-Histogram Equalization (MMBEBHE)

(1) (2)

In MMBEBHE [14], Mean Brightness Error (MBE) of the original image is calculated and the threshold level which yield minimum MBE has to be calculated. Then separate the image into two sub images based on the threshold value and the equalization has to be performed.

Let ET(Y) denote the output mean of the BBHE with threshold level set as XT.

Assuming X0=0 and 1+Xl-1=L ,

Et (Y) =1 k + L(1 "XLo P(X ))

The output mean with threshold level set as XT+1.

Et+i(Y ) = ET (Y) + 2 [l - LP ( Xt+i)] (5)

Mean Brightness Error (MBE),

MBE = E(Y) - E(X) (6)

where E(X) is input mean, E(Y) is output mean. 3.5 Recursive Mean-Separate Histogram Equalization (RMSHE)

Separating the mean before performing histogram equalization provides better contrast enhancement with brightness preservation [15, 22].Chen et al proposed RMSHE in which image is separated into two sub images based on the mean of original image. After separating the mean, the histogram of the two

images are equalized and the mean separation is done recursively. More mean separation gives more brightness preserved contrast enhancement.

Fig.2. (a) - (e) Contrast enhancement of HE, CLAHE, BBHE, MMBEBHE, RMSHE techniques, (f) - (j) Histograms of the corresponding techniques.

Table 1. EME values for different contrast enhancement techniques

Image HE CLAHE BBHE MMBE BHE RMSHE

mdb03 mdb06 mdb09 mdb10 mdb12 mdb15 mdb17 mdb19 1.075 1.758 2.300 0.797 1.338 0.622 0.373 0.834 6.950 4.730 10.242 5.469 6.017 6.084 4.690 6.792 1.863 3.077 3.412 3.068 2.866 2.293 0.905 1.305 2.718 5.579 5.924 3.889 4.742 3.983 1.347 2.343 7.519 4.616 8.413 5.528 5.681 7.595 5.727 6.697

Table 2. PSNR values for different contrast enhancement techniques

Image HE CLAHE BBHE MMBE BHE RMSHE

mdb03 mdb06 mdb09 mdb10 mdb12 mdb15 mdb17 mdb19

6.60 8.59 8.15

4.61 9.76 4.39 3.52

23.68 22.80 24.67 25.35

24.72 24.87

26.73 22.17

17.98 15.08 18.28 18.25 18.59 18.75 21.17 13.72

32.07 31.60 32.54 25.85

27.24 23.67 24.00 27.39 26.94 24.06 33.89 22.04

For recursion level r = n, the output mean E(Y) is as follows [15]:

E(Y) = Xm +

where XG is the middle gray level and Xm is the input mean. As the recursion level, n increases, E(Y) will eventually converge to the input mean which is evident from the above equation.

( Xq - Xm ).

4 Performance Evaluation

The goal of the enhancement algorithm is to improve the image quality so that the processed image is better than the original image for further processing. Such an enhancement can be assessed subjectively by a visual inspection of the image. However, a precise and complete characterization cannot be achieved by subjective evaluation. There is no universal measure that can specify both the objective and the subjective validity of the algorithm. However in this paper Effective Measure of Enhancement (EME) and Peak Signal to Noise Ratio (PSNR) are used to evaluate the performance of the algorithms. PSNR is a measure of the deviation of the current image from the original image with respect to the peak value of the gray level. The EME is a quantitative measure of image enhancement. It is obtained by splitting the image into a number of blocks and using the equation,

K2 K1 f T ilr 1\\

K1K 2 L=1 K=1

Imax(k , l) V Imin (k.l)

where,Kj,K2 are the number of horizontal and vertical blocks in the image, Imax(k,1) and Imm(k,1) are the maximum and minimum pixel values in a given block [16].

Table.1 shows the EME measure values calculated using equation (8) for the few images from MIAS database. The results show that the CLAHE and RMSHE provide better enhancement compared to the other methods.

While inspecting the images with human eyes, enhanced pixels by CLAHE mimics micro calcifications as reported in [9].The detailed information of the image is enhanced better by MMBEBHE as shown in Fig.2 (d). But the boundary of the image is merged with the back ground. Table.2 suggests that MMBEBHE provides better noise free image compared to other methods. CLAHE and RMSHE perform equally good results with less noise.

5 Conclusion

In this paper, the indirect contrast enhancement techniques HE, CLAHE, BBHE, MMBEBHE, and RMSHE are compared by applying them to the low contrast mammographic images. The traditional HE method significantly changes the brightness of the image. Hence the details of the image cannot be viewed clearly. MMBEBHE provides better noise removal in the images. By applying some edge preservation filter to it, the performance of enhancement can be increased.

Both CLAHE and RMSHE techniques are found to offer better enhancements of masses and micro calcifications present in the images taken for study. However, based on the detailed performance evaluation studies conducted in this work, it is clearly evident that the RMSHE technique achieves the best contrast enhancement for the low contrast mammographic images taken for study and it offers the best brightness preservation as well.

References

[1] Tang, Jinshan, Rangaraj M. Rangayyan, Jun Xu, Issam El Naqa, and Yongyi Yang. Computer-aided detection and diagnosis of breast cancer with mammography: recent advances. Information Technology in Biomedicine, IEEE Transactions on 13,

no. 2, 2009: 236-251.

[2] Morrow, William Mark, Raman Bhalachandra Paranjape, Rangaraj M. Rangayyan, and Joseph Edward Leo Desautels. Region-based contrast enhancement of mammograms. Medical Imaging, IEEE Transactions on 11, no. 3, 1992: 392-406.

[3] Cheng, Heng-Da, and Huijuan Xu. A novel fuzzy logic approach to contrast enhancement. Pattern Recognition 33, no.

5,2000: 809-819.

[4] Rangayyan, Rangaraj M., Liang Shen, Yiping Shen, JE Leo Desautels, Heather Bryant, Timothy J. Terry, Natalka Horeczko, and M. Sarah Rose. Improvement of sensitivity of breast cancer diagnosis with adaptive neighborhood contrast enhancement of mammograms. Information Technology in Biomedicine, IEEE Transactions on 1, no. 3,1997: 161-170.

[5] Tang, Jinshan, Xiaoming Liu, and Qingling Sun.A direct image contrast enhancement algorithm in the wavelet domain for screening mammograms.Selected Topics in Signal Processing, IEEE Journal of 3, no. 1 ,2009: 74-80.

[6] Zhou, Yicong, Karen Panetta, and Sos Agaian. Human visual system based mammogram enhancement and analysis. In Image Processing Theory Tools and Applications (IPTA), 2010 2nd International Conference on, pp. 229-234. IEEE, 2010.

[7] Pisano, Etta D., Elodia B. Cole, Bradley M. Hemminger, Martin J. Yaffe, Stephen R. Aylward, Andrew DA Maidment, R. Eugene Johnston et al. Image Processing Algorithms for Digital Mammography: A Pictorial Essay 1.Radiographics 20, no. 5,2000: 1479-1491.

[8] Pisano, Etta D., Shuquan Zong, Bradley M. Hemminger, Marla DeLuca, R. Eugene Johnston, Keith Muller, M. Patricia Braeuning, and Stephen M. Pizer. Contrast limited adaptive histogram equalization image processing to improve the detection of simulated spiculations in dense mammograms. Journal of Digital Imaging 11, no. 4,1998: 193-200.

[9] Sivaramakrishna, Radhika, Nancy A. Obuchowski, William A. Chilcote, Gilda Cardenosa, and Kimerly A. Powell.Comparing the performance of mammographic enhancement algorithms: a preference study. American Journal of Roentgenology 175, no. 1,2000: 45-51.

[10] Heinlein, Peter, Johann Drexl, and Wilfried Schneider.Integrated wavelets for enhancement of microcalcifications in digital mammography. Medical Imaging, IEEE Transactions on 22, no. 3,2003: 402-413.

[11] Mencattini, Arianna, Marcello Salmeri, Roberto Lojacono, Marco Frigerio, and Federica Caselli. Mammographic images enhancement and denoising for breast cancer detection using dyadic wavelet processing. Instrumentation and Measurement, IEEE Transactions on 57, no. 7,2008: 1422-1430.

[12] Krishnamoorthy, R., N. Amudhavalli, and M. K. Sivakkolunthu. An Adaptive Mammographic Image Enhancement in Orthogonal Polynomials Domain.International Journal of Computer and Information Engineering 4, no. 2,(2010: 120-128.

[13] Kim, Yeong-Taeg. Contrast enhancement using brightness preserving bi-histogram equalization.Consumer Electronics, IEEE Transactions on 43, no. 1, 1997: 1-8.

[14] Chen, Soong-Der, and Abd Rahman Ramli. Minimum mean brightness error bi-histogram equalization in contrast enhancement. Consumer Electronics, IEEE Transactions on 49, no. 4,2003: 1310-1319.

[15] Chen, Soong-Der, and Abd Rahman Ramli. Contrast enhancement using recursive mean-separate histogram equalization for scalable brightness preservation. Consumer Electronics, IEEE Transactions on 49, no. 4,2003: 1301-1309.

[16] Agaian, Sos S., Karen Panetta, and Artyom M. Grigoryan. Transform-based image enhancement algorithms with performance measure. Image Processing, IEEE Transactions on 10, no. 3,(2001: 367-382.

[17] Wang, Chao, and Zhongfu Ye. Brightness preserving histogram equalization with maximum entropy: a variational perspective. Consumer Electronics, IEEE Transactions on 51, no. 4 ,2005: 1326-1334.

[18] Abdullah-Al-Wadud, Mohammad, Md Hasanul Kabir, M. Ali Akber Dewan, and Oksam Chae. A dynamic histogram equalization for image contrast enhancement. Consumer Electronics, IEEE Transactions on 53, no. 2,(2007: 593-600.

[19] Sundaram, M., K. Ramar, N. Arumugam, and G. Prabin."Histogram based contrast enhancement for mammogram images. In Signal Processing, Communication, Computing and Networking Technologies (ICSCCN), 2011 International Conference on, pp. 842-846. IEEE, 2011.

[20] Lu, Li, Yicong Zhou, Karen Panetta, and Sos Agaian. Comparative study of histogram equalization algorithms for image enhancement. In SPIE Defense, Security, and Sensing. International Society for Optics and Photonics, 2010.

[21] Zhao, Yisu, Nicolas D. Georganas, and Emil M. Petriu. Applying contrast-limited adaptive histogram equalization and integral projection for facial feature enhancement and detection. In Instrumentation and Measurement Technology Conference (I2MTC), 2010 IEEE, pp. 861-866.

[22] Lee, Eunsung, Sangjin Kim, Wonseok Kang, Doochun Seo, and Joonki Paik. Contrast enhancement using dominant brightness level analysis and adaptive intensity transformation for remote sensing images. Geoscience and Remote Sensing Letters, IEEE 10, no. 1,2013: 62-66.

[23] http://www.mammoimage.org/databases/

[24] www.cancer.gov/