Denoising and compression of glaucomatous fundus image using wavelet transform

(1)

International Journal of Recent Advances in Engineering & Technology (IJRAET)

________________________________________________________________________________________________

ISSN (Online): 2347 - 2812, Volume-3, Issue -9, 2015 58

Denoising and compression of glaucomatous fundus image using wavelet transform

1Sima Sahu, ²Harsh Vikram Singh, ³Basant Kumar

1Dept. of ECE, MRCE, Maisammaguda, Dhulapally, Secunderabad, India,

2Dept. of Electronics, KNIT, Sultanpur, U.P., India 3 Dept. of ECE, MNNIT, Allahabad, U.P., India

Abstract— Glaucoma is one of the major causes of preventable blindness in the world. It is often associated with the increase the pressure (IOP) of the fluid in the eye, and it is called “Silent Thief of Sight”. It is presently detected either by regular inspection of the retina, measurement of the intra ocular pressure (IOP) or by a loss of vision. But present manual diagnosis is expensive, tedious and time consuming. So detection of glaucoma using fundus retinal image is more promising and superior.

Medical Images normally have a problem of high level components of noises. Image denoising is an important task in image processing. Use of wavelet transform improves the quality of an image and reduces noise level. The paper deals with the use of wavelet transform for image de- noising by employing a selected method of thresholding of appropriate decomposition coefﬁcients.

Keywords-Glaucoma, Wavelet, Image de-noising, Medical image

I. INTRODUCTION

GLAUCOMA is a chronic eye disease in which the optic nerve is progressively damaged. It is the second leading cause of blindness, and is predicted to affect around 80 million people by 2020 [1]. Progression of the disease leads to loss of vision, which occurs gradually over a long period of time. As the symptoms only occur when the disease is quite advanced, glaucoma is called the silent thief of sight. Glaucoma cannot be cured, but its progression can be slowed down by treatment.

Therefore, detecting glaucoma in time is critical.

Image processing field is a huge one. It encompasses in the following areas: 1. Image Compression; 2. Image De-noising; 3. Image Enhancement; 4. Image Recognition; 5. Feature Detection, 6. Texture Classification. Wavelet-based techniques apply to all of these topics. Reason is wavelet analysis provides such an all-encompassing tool for image processing for a similar type of analysis occurs in the human visual system. To be more precise, the human visual system performs hierarchical edge detection at multiple levels of resolution and wavelet transforms perform a similar analysis. Compression is the process of reducing large data files into smaller files for efficiency of storage and transmission. Data compression techniques are: a.

Lossless data compression, b. Lossy data compression.

Lossless data compression is nothing but the original data can be reconstructed exactly from compressed data.

Lossy data compression in which data after compression and then decompression retrieves a file that is not exactly as the original data as there will be loss of data.

De-noising plays a very important role in the field of the medical image pre-processing.

It is often done before the image data is to be analyzed.

Denoising is mainly used to remove the noise that is present and retains the significant information, regardless of the frequency contents of the signal. It is entirely different content and retains low frequency content. De-noising has to be performed to recover the useful information. In this process much attention is kept on, how well the edges are preserved and how much of the noise granularity has been removed [3]. The main purpose of an image-denoising algorithm is to eliminate the unwanted noise level while preserving the important features of an image. In wavelet domain, the noise is uniformly spread throughout the coefficients while mostly the image information is concentrated in the few largest coefficients..

II. WAVELET TRANSFORM

In most of the applications of image processing, it is essential to analyze a digital signal. If the data will be transformed into any other domain then the structure and features of the signal may be better understood. There are several transforms available like Fourier transform, Hilbert transform, Wavelet transform, etc. The wavelet transform is better than Fourier transform because it gives frequency representation of raw signal at any given interval of time, but Fourier transform gives only the frequency- amplitude representation of the raw signal but the time information is lost [5]. So we cannot use the Fourier transform where we need time as well as frequency information at the same time. Wavelets are mathematical functions that cut up data into different frequency components. The fundamental idea behind wavelets is to analyze the signal at different scales or resolutions, which is called multiresolution. The most

(2)

________________________________________________________________________________________________

important feature of wavelet transform is it allows multiresolution decomposition.

III. IMAGE DENOISING

An image is often corrupted by noise in its acquition and transmission. Image de-noising is used to remove the additive noise while retaining as much as possible the important signal features. In the recent years there has been a fair amount of research on wavelet thresholding and threshold selection for signal de-noising [14], because wavelet provides an appropriate basis for separating noisy signal from the image signal. The motivation is that as the wavelet transform is good at energy compaction, the small coefficients are more likely due to noise and large coefficient due to important signal features. These small coefficients can be thresholded without affecting the significant features of the image.

IV. BASIC MODEL OF COMPRESSION SYSTEM

Most of the compression systems are based upon reducing the redundant information present in the signal whether it is 1D signal or 2D signal like image.

Sometime redundancy reduction process is performed over the transformed signal rather than the original signal itself. The redundancy depends upon the entropy of the signal. Redundancy reduction removes highly correlated data which is more in case of image due to much of the low frequency content in it. Discrete Wavelet Transform (DWT) [12] has emerged as a popular technique for redundancy reduction. DWT has high de-correlation and energy compaction efficiency .Non-significant information is removed from the data by this process but it is a non-reversible process electronically [13] The popular entropy coding techniques are Huffman coding and Arithmetic coding which are used many times in compressing data and also in image compression techniques. An image that is decomposed by wavelet transform can be reconstructed with desired resolution. The most important feature of wavelet transform is it allows multiresolution decomposition. The procedure for this is a low pass filter and a high pass filter is chosen, such that they exactly halve the frequency range between themselves.

This filter pair is called the Analysis Filter pair. First of all, the low pass filter is applied for each row of data, and then we obtain low frequency components of the row. As the LPF is a half band filter, the output data consists of frequencies only in the first half of the original frequency range. By Shannon's Sampling Theorem they can be sub sampled by two, so that the output data contains only half the original number of samples, similarly the high pass filter is applied for the same row of data, and now the high pass components are separated, and placed by the side of the low pass components. This procedure is done for all rows. The different levels of Wavelet decomposition is shown in the following Fig 1.

LL HL

LH HH

1st level

LL HL HL LH HH

LH HH

2nd level

LL HL HL HL LH HH

LH HH

3rd level

Figure 1: wavelet decomposition

Next, the filtering is done on each column. as a result we get four bands of data, each labeled as LL (low-low), HL (high-low), LH (low-high) and HH (high-high).The LL band can be decomposed once again in the same manner, thereby producing even more sub bands[8].

This can be done up to any level, thereby resulting in a pyramidal decomposition as shown above the LL band at the highest level can be said as most important, and the other bands are of lesser importance, the degree of importance decreases from the top of the pyramid to bottom. The following figure :2 shows the block diagram for image compression.

Figure 2: Block diagram for image compression

V. EXPERIMENTAL RESULTS

The following retinal fundus image shown in Fig 3 is used for denoising and compression using Wavelet 2D transform.

MATLAB 7.0 version is used for simulation. The output results after denoising and compression is shown in the following figures. First of all the original image is decomposed by using wavelet decomposition as shown in the following Fig 4. In this case 2^nd level decomposition is used. Next the decomposed image is de-noised by using soft thresholding and haar wavelet method as shown in the Fig 5. Here the threshold level is kept 1.548 for level 1 and 1.794 for level 2.The noise structure is taken as the unscaled white noise.The de- noised image along with the original image is shown in the following Fig 6.

Next By using wavelet 2D the image is compressed.

First of all the original image is analyzed at level 2 with

Image Forward

wavelet transform

Quantizer

Retained image

Inverse wavelet transform

Encode

Dequantizer Decode

(3)

________________________________________________________________________________________________

haar wavelet. The analyzed image is shown in Fig 7.The retained energy is 99.71% and number of zeros is 93.75%. Here balance sparsity-norm thresholding method is used with global threshold value 106 for compression. The original image with the compressed image is shown in Fig 8.

The performance evaluation of wavelet transform can be found out by comparing it with other filters. The performance results are evaluated according to their error criteria namely MSE (mean square error) and PSNR (Peak signal to noise ratio).The comparison is shown in the following table 1. From this table wavelet transform shows the good performance over all other filters.

The results tell what percentage of the wavelet coefficients was set to zero and what percentage of the image’s energy is preserved in the compression process.

Here even though the compressed image is constructed from only about half as many nonzero wavelet coefficients as the original, there is almost no detectable deterioration in the image quality.

Figure 3: Original image

Figure 4: Analyzed image

Figure 5: De-Noising

Figure 6: De-Noised Image

Figure 7: compression

Figure 8: compressed image

TABLE 1. : (MSE and PSNR values for different types of filter)

(4)

________________________________________________________________________________________________

Filter MSE PSNR Mean 99.7794 28.151 Median 80.9181 29.0738 Gaussian 87.844 28.7055 Weiner 107.0784 27.8382 Wavelet 75.9231 31.0981

VI. CONCLUSION AND FUTURE

In this paper we have taken a fundus retinal image for de-noising and compression in Wavelet Toolbox specially Wavelet 2D in MATLAB. As a result we get the compressed image with retained energy 99.71% as well as noise free in vertical, horizontal and diagonal details. From the performance parameter shown in Table 1.It can be concluded that wavelet transform is an optimum filter that can be best to apply to all fundus retinal image. So there is almost no detectable deterioration in the image quality.

In future work, we can denoise a fundus retinal image using different wavelets. We can do that making a different wavelet dimension to get more and more noise free and compressed image. Here we have gone through level 2 decomposition, in future we can make it to level 3, 4 or more to be more accurate and clear.

REFERENCES

[1] Jun Cheng, Jiang Liu, Yanwu Xu, Fengshou Yin, Damon Wing Kee Wong, Ngan-Meng Tan, Dacheng Tao, Ching-Yu Cheng, Tin Aung, and Tien Yin Wong.: Superpixel Classification Based Optic Disc and Optic cup Segmentation for Glaucoma Screening. IEEE TRANSACTIONS

ON MEDICAL IMAGING, VOL.32,

NO.6,JUNE 2013.

[2] Meier, J., Bock, R., Michelson, G., Ny´ul, L.G., Hornegger, J.: Effects of prepro-cessing eye fundus images on appearance based glaucoma classification. In: Procceedings of International Conference on Computer Analysis of Images and Patterns (2007) (accepted fo r publication).

[3] Narasimha-Iyer, H., Can, A., Roysam, B., Stewart, C.V., Tanenbaum, H.L., Ma-jerovics, A., Singh, H.: Robust detection and classification of longitudinal changesin color retinal fundus images for monitoring diabetic retinopathy. IEEE Trans.Biomed. Eng. 53(6), 1084–1098 (2006) . [4] Lester, M., Garway-Heath, D., Lemij, H.: Optic

Nerve Head and Retinal NerveFibre Analysis.

European Glaucoma Society (2005).

[5] Chr´astek, R., Wolf, M., Donath, K., Niemann, H., Paulus, D., Hothorn, T., Lausen,B.,L¨ammer, R., Mardin, C., Michelson, G.: Automated segmentation of the opticnerve head for diagnosis of glaucoma. Med. Image Anal. 9(4), 297– 314 (2005)

[6] Zangwill, L.M., Chan, K., Bowd, C., Hao, J., Lee, T.W., Weinreb, R.N., Sejnowski, T.J.,Goldbaum, M.H.: Heidelberg retina tomograph measurements of the opticdisc and parapapillary retina for detecting glaucoma analyzed by machine learning classifiers. Invest.

Ophthalmol. Vis. Sci. 45(9), 3144–3151 (2004).

[7] Adler, W., Hothorn, T., Lausen, B.: Simulation based analysis of automated, clas-sification of medical images. Methods Inf. Med. 43(2), 150–

155 (2004).

[8] Staal, J., Abr`amoff, M., Niemeijer, M., Viergever, M., van Ginneken, B.: Ridge-based vessel segmentation in color images of the retina.

IEEE Trans. Med.Imag. 23(4), 501–509 (2004) [9] Zhao, W., Chellappa, R., Phillips, P.J.,

Rosenfeld, A.: Face recognition: A literaturesurvey. ACM Comput. Surv. 35(4), 399–458 (2003) .

[10] Greaney, M.J., Hoffman, D.C., Garway-Heath, D.F., Nakla, M., Coleman, A.L.,Caprioli, J.:

Comparison of optic nerve imaging methods to distinguish normal eyesfrom those with glaucoma. Invest. Ophthalmol. Vis. Sci. 43(1), 140–145 (2002).

[11] Hoover, A., Goldbaum, M.: Locating the optic nerve in a retinal image using thefuzzy convergence of the blood vessels. IEEE Trans.

Med. Imag. 22(8), 951–958(2003)

[12] Martin Vetterli S Grace Chang, Bin Yu.:

Adaptive wavelet thresholding for image denoising and compression. IEEE Transactions on Image Processing,9(9):1532–1546, Sep 2000.

[13] Swindale, N.V., Stjepanovic, G., Chin, A., Mikelberg, F.S.: Automated analysis ofnormal and glaucomatous optic nerve head topography images. Invest. Ophthalmol.Vis. Sci. 41(7), 1730–1742 (2000).

[14] Iester, M., Swindale, N.V., Mikelberg, F.S.:

Sector-based analysis of optic nervehead shape parameters and visual field indices in healthy and glaucomatous eyes.J. Glaucoma. 6(6), 370–376 (1997).

[15] Uchida, H., Brigatti, L., Caprioli, J.: Detection of structural damage from glaucomawith confocal laser image analysis. Invest. Ophthalmol. Vis.

Sci. 37(12), 2393–2401(1996).

[16] Malinovsky, V.E.: An overview of the Heidelberg Retina Tomograph. J. Am. Op-tom.

Assoc. 67(8), 457–467 (1996)

[17] Sivalingam, E.: Glaucoma: An overview. J.

Ophthalmic. Nurs. Tech. 15(1), 15–18(1996)

