Biomedical Image Enhancement Using Wavelets

Available online at www.sciencedirect.com

ScienceDirect Procedia Computer Science 48 (2015) 513 – 517

International Conference on Intelligent Computing, Communication & Convergence (ICCC-2014) (ICCC-2015) Conference Organized by Interscience Institute of Management and Technology, Bhubaneswar, Odisha, India

Biomedical Image Enhancement Using Wavelets Kirti Khatkara,*, Dinesh Kumarb Research Scholar,CSE Dept.GJUS&T, Hisar, Haryana,India

a b

CSE Dept.GJUS&T, Hisar, Haryana, India

Abstract

As image enhancement is the main issue for biomedical image diagnosis, the aim of this paper is to present a method to enhance the biomedical images. In this paper, a combination of wavelets is used for the same. In the method after applying SIFT(Scale Invariant Feature Transforms) algorithm on the image the first wavelet D’Mayer is applied, then image is extracted and the second wavelet Coieflet is applied on the image. The results of the proposed method have been compared with other wavelets on the basis of different metrics like PSNR (Peak signal to noise ratio) and Beta coefficient and it has been found that the proposed method provides better results than the other methods. Keywords: Image enhancement; biomedical imaging; wavelet transform 1. Introduction Image enhancement is designed to improve the picture quality of an image. Various methods like Filtering, Wavelet Transformation and Soft Computing techniques are used for biomedical image enhancement. Filtering is the process of removing unwanted components from a signal. As non-linear filters like Median filters [9], Wiener filters [15] provide better results by providing better edge preservation and good PSNR values so these are mostly used for biomedical image denoising. Various other filters like unsharp masking [12], digital unsharp masking [7] and spatial band pass filtering [13] have also been used for the contrast enhancement of mammographic images. Soft computing *

Corresponding author email id:[email protected] (K.Khatkar), [email protected] (D. Kumar)

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 scientific committee of International Conference on Computer, Communication and Convergence (ICCC 2015) doi:10.1016/j.procs.2015.04.128

514

Kirti Khatkar and Dinesh Kumar / Procedia Computer Science 48 (2015) 513 – 517

techniques are applied to match the spectra to the type of problem [11]. Some of the soft computing techniques include Neural Networks, Fuzzy Logic and Genetic Algorithm. Use of Neural networks for the contrast enhancement of images that includes colored images like stained media for TB bacilli has been proved an important tool for diagnostic purpose as the visibility is increased [2]. Fuzzy filters are also capable of dealing with the images that are highly affected due to noise [10]. Fuzzy Logic is constructive for image enhancement, image segmentation, image classification and thresholding value selection. Genetic Algorithm is a technique of breeding computer programs and solutions to optimization by taking simulated evolution that helps in contrast enhancement and evaluating detailed structure of image [5]. 2. Biomedical Image Denoising using Wavelets The main advantage of Wavelet transform is that it is capable of proving localization in space as well as frequency domains. The wavelet transform is more reliable and it can provide the exceptional information for dissimilar resolutions [14]. The use of non-linear mapping functions derived for projecting a set of discrete wavelet transform (DWT) provides better enhancement in medical images in comparison to fast Fourier transform (FFT) and the conventional wavelet based methods [4]. On the other hand, a multiwavelet system enhance the image by simultaneously providing perfect reconstruction while preserving length (orthogonality), good performance at the boundaries (via linear-phase symmetry) and a high order of approximation (vanishing moments) [6]. 3. Proposed method In the proposed method, a combination of wavelets has been used for medical image enhancement. After applying SIFT algorithm, first wavelet D’Mayer is applied on the image. Then image boundaries are extracted and the second wavelet Coieflet is applied on the image. Figure 1 shows the flowchart of the image enhancement using the proposed 2-D wavelet coefficient mapping. The flowchart of proposed method is Start Upload image

Key point selection using SIFT algorithm

Apply D’Mayer wavelet on the image

Extract the image boundaries Apply Coieflet wavelet on the extracted image

Calculate PSNR and BETA values for the extracted image

Compare the results with other wavelets Fig. 1. Flowchart of proposed method

Step1: The SIFT algorithm is used for key point selection. SIFT improves the key points and throws out the bad ones i.e. the filtering and localization of the key points. It creates the descriptor using histograms of orientations. SIFT finds the scale space extrema, key point, localization, orientation assignment, key point descriptor [3]. After applying SIFT algorithm the results are shown in figure 2

Fig. 2. Key points after applying SIFT algorithm

Step2: On the resulting images of SIFT algorithm, the D’Mayer wavelet transformation is applied. The images after D’Mayer wavelet transformation is shown in figure 3

515

Kirti Khatkar and Dinesh Kumar / Procedia Computer Science 48 (2015) 513 – 517

Fig. 3. Results after applying D’Mayer algorithm

Step3: The resulting image boundaries are extracted after applying D’Mayer wavelet transformations shown in figure 4

Fig.4

Step 4: After extracting the image the second wavelet Coieflet is applied on the image. The resulting images are shown in figure 5.

Fig.5 Resulting images

4. Metrics for Comparison The table describe comparison of different metrics like PSNR and Beta values with the other wavelets for extracted images. PSNR is the ratio between the maximum possible power of a signal and the power of corrupting noise that affects the reliability of its representation [8]. PSNR is described as PSNR = 10 log10 (

=20 log10 (

)

(1)

The beta metric is used as edge preservation measure in the filtered image [1]

β=

;

(I1,I2)=

(2)

where ∆I and ∆Î represent the high pass filtered version of original image I(i,j) and its denoising version Î(i,j). and are the mean intensities of ∆I and ∆Î respectively. An increasing indicates a better image quality. 5. Experimental Results The proposed method is applied on various high quality ultrasound images and results are shown in the tabular form. Various images particularly Heart, Liver, Kidney and Brain images are selected as shown in figure 6

Fig. 6:Standard images of Heart,Brain,Kidney and liver used in the proposed method

The comparision of the proposed method with the other existing wavelets like Coieflet, Daabuchiee, D’ Mayer, Symlet for different images like Heart, Brain, Kidney and Liver for average value of ten iterations has been given in the below tables as under Table 1: Different values of PSNR and BETA coefficient for Brain Noise variance

D’ Mayer PSNR

10 20 30 40 50

55.7151 55.6795 55.6658 55.6885 55.6647

BETA 1.1079 1.1054 1.151 1.1515 1.1158

Coieflet PSNR 55.7156 55.6895 55.6768 55.6994 55.6774

BETA 1.1046 1.1107 1.0826 1.1335 1.1194

Symlet PSNR 55.7156 55.6894 55.6769 55.6994 55.6774

BETA 1.1391 1.1315 1.0924 1.1278 1.1235

Daabuchiee PSNR BETA

Proposed method PSNR BETA

55.7152 55.6797 55.6657 55.6887 55.6649

83.5727 83.5284 83.4988 83.5329 83.4971

1.1178 1.1284 1.0861 1.1611 1.0935

2.1088 2.1432 2.1409 2.1141 2.1375

516

Kirti Khatkar and Dinesh Kumar / Procedia Computer Science 48 (2015) 513 – 517

Table 2: Different values of PSNR and BETA coefficient for Kidney Noise variance

D’ Mayer PSNR

10 20 30 40 50

52.8646 53.1754 52.8942 52.8820 52.8819

BETA 1.0738 1.1072 1.1311 1.0797 1.0708

Coieflet PSNR 52.9014 53.1913 52.9364 52.9199 52.9247

BETA 1.0860 1.1052 1.1128 1.1345 1.1355

Symlet PSNR 52.9015 53.1915 52.9364 52.8967 52.9256

Daabuchiee PSNR BETA

Proposed method PSNR BETA

1.1151 1.1128 1.0726 1.0973 1.1127

52.8640 53.1756 52.8942 52.8611 52.8817

79.2968 79.7635 79.3412 79.2915 79.3226

BETA

1.1371 1.1120 1.0748 1.0887 1.1344

2.1047 2.0937 2.0861 2.1224 2.1096

Table 3:Different values of PSNR and BETA coefficient for liver Noise variance

D’ Mayer PSNR BETA

Coieflet PSNR

BETA

Symlet PSNR

BETA

Daabuchiee PSNR BETA

Proposed method PSNR BETA

10 20 30 40 50

57.5185 57.5085 57.5165 57.5054 57.5106

57.5619 57.5491 57.5619 57.5462 57.5511

1.1151 1.0962 1.1438 1.1221 1.0874

57.5617 57.5489 57.5619 57.5464 57.5511

1.1134 1.1471 1.1332 1.1307 1.1181

57.5188 57.5085 57.5165 57.5055 57.5111

86.2782 86.2628 86.2747 86.2576 86.2654

1.1294 1.0929 1.0998 1.0783 1.0726

1.1478 1.1493 1.1618 1.1150 1.1365

2.1294 2.1239 2.0758 2.0739 2.1492

Table 4: Different values of PSNR and BETA coefficient for Heart Noise variance

D’ Mayer PSNR BETA

Coieflet PSNR

BETA

Symlet PSNR

BETA

Daabuchiee PSNR BETA

Proposed method PSNR BETA

10 20 30 40 50

59.3119 59.2953 59.3139 59.3111 59.2988

59.3881 59.3634 59.3828 59.3862 59.3727

1.1032 1.1509 1.1032 1.1307 1.1468

59.3881 59.3635 59.3828 59.3864 59.3727

1.0838 1.1105 1.1234 1.0723 1.1366

59.3119 59.2957 59.3139 59.3111 59.2989

88.9678 88.9433 88.9708 88.9665 88.9481

1.0817 1.1261 1.0937 1.0871 1.1229

1.1305 1.1029 1.0969 1.1566 1.1334

2.1128 2.1119 2.1211 2.1157. 2.0836

6. Conclusion In this paper, a comparative study of the performance of different wavelets with the proposed method has been presented. The PSNR value and the BETA coefficient for different wavelets like Symlet, Daabuchiee, Coieflet and D’Mayer are compared with the proposed method and it has been found that the proposed method provides better results than other wavelets. Four biomedical images brain, kidney, heart and liver have been considered and the results are obtained by providing different noise levels. As the proposed method yields better value of Beta, which demonstrates clearer boundary or edge values of the biomedical images that are affected due to noise or other factors. References 1. Adamo, GregorioAndria, Filippo, Attivissimo, Anna Maria Lucia Lanzolla, Maurizio Spadavecchia. A Comparative study on mother wavelet selection in ultrasound image denoising. Measurement, Elsevier; 2013; 46:2447-56. 2. A.S.W. Wahab,M.Y. Mashor, Zaleha Salleh, S. A. Abdul Shukor, N. Abdul Rahim,F.Muhamad Idris, H. Hasan, S.S. Md Noor. A Neural Network Approach for Contrast Enhancement Image. International Conference on Electronic Design; 1-3 Dec.,2008; Penang, Malaysia. 3. David G. Lowe. Distinctive Image Features from Scale-Invariant Key points. International Journal of Computer Vision; 2004. 4. Du-Yih Tsai and Yongbum Lee. A method of medical image enhancement using Wavelet-Coefficient Mapping Functions.IEEE International. Conf. Neural Networks & Signal Processing Nanjing, China; 2003.p. 14-17. 5. G. Farias, M. Santos. Making decisions on Brain Tumor diagnosis by Soft Computing Techniques. Soft Comput., Springer; 2010; 14:1287–96. 6. Hai-Hui Wang, Jun Wang, Wei Wang. Multispectral Image Fusion Approach Based on GHM Multi Wavelet Transform. Proc.of the Fourth International Conference on Machine Learning and Cybernetics; 2005; 8:18-21. 7. H. P. Chan, C. J. Vybomy, H. MacMahon, C. E. Metz, K. Doi, and E. A. Sickles. Digital Mammography: ROC studies of the effects of pixel size and unsharp-mask filtering on the detection of subtle microialcifications. Investigative Radiology; 1987; 22(7):581-89. 8. Huynh-Thu,Q. Ghanbari.Scope of Validity of PSNR in image/video quality Assessment.IEEE;2008;44(13):800–1. 9. J.W.Tukey.Nonlinear (nonsuperposable) Methods for smoothing data. Proc.Congr. Rec.EASCOM; 1974.p.673. 10. Koushik Mondal, Paramartha Dutta, Siddhartha Bhattacharyya. Gray Image extraction using Fuzzy Logic. Second International Conference on Advanced Computing & Communication Technologies, IEEE; 2012.p.289-296. 11. Mantas Paulinas, Andrius Ušinskas. A Survey of Genetic Algorithms Applications for Image Enhancement and Segmentation.Information Technology and Control;2007;36(3):278-84. 12. M.B.McSweeney, P.S.Prawls and R.L. Egan. Enhanced image mammography. AJR; 1983;140: 9-14. 13. R.L.Smathers, E.Bush, J.Drace, M. Stevens, F.G.Sommer, B.W.Brownand B. Kanas.Mammographic micro calcifications: Detection with

Kirti Khatkar and Dinesh Kumar / Procedia Computer Science 48 (2015) 513 – 517 xerography, screen-film and digitized film display, Radiology; 1986; 159(3):673-77. 14. Vasily Strela, Peter Niels Heller. The Application of Multiwavelet Filter banks to Image Processing. IEEE Trans. On Image Processing; 1997; 8(4):548-63. 15. Yoshikazu Washizawa, Yukihiko Yamashita.Non-linear Wiener filter in reproducing kernel Hilbert space. Proceedings of the 18th International Conference on Pattern Recognition (ICPR'06); 2006; 1:967-70.

517