Vo lum e 1, N o .

9, N o vem ber 2012 ISSN 2278-1080 Th e In tern atio n al Jo urn al o f C o m puter Scien ce & Application s (TIJCSA) RESEARCH PAPER Available O n lin e at h ttp://w w w .journ alofcom puterscien ce.com /

Adaptive Fuzzy Approach to Gaussian Noise Removal in Gray Scale Images

Deepinder Kaur1
Department of Computer Science & Engineering, B.B.S.B.E.C, Fatehgarh Sahib(Punjab), India [email protected] Baljit Singh2 Department of Computer Science & Engineering, B.B.S.B.E.C, Fatehgarh Sahib(Punjab), India [email protected]

Abstract
Visual information transmitted in the form of digital images is becoming a major method of communication in the modern age, but the image obtained after transmission is often corrupted with noise. The received image needs processing before it can be used in applications. A New Fuzzy Filter that adopts Fuzzy Logic is proposed in this paper which removes Gaussian Noise from the Corrupted Gray scale Images. The main concern of the present filter is to distinguish between local variations due to noise and due to image structure. It uses 14 fuzzy rule based convolution mask on every pixel of the image. Objective performance of the proposed algorithm is compared with conventional methods based on Mean Square Error (MSE), Root Mean Square Error (RMSE), Signal to Noise Ratio (SNR) and Peak Signal to Noise Ratio(PSNR). The results illustrate that the proposed method can be used as an effective Noise removal method for
2012, http://www.journalofcomputerscience.com - TIJCSA All Rights Reserved

Deepinder Kaur ,Baljit Singh, The International Journal of Computer Science & Applications (TIJCSA) ISSN 2278-1080, Vol. 1 No. 9 November 2012

Gaussian noise. Results show that by using the proposed method improved SNR and PSNR and minimized MSE and RMSE are achieved. Hence proposed algorithm leads to better image enhancement.

Keywords: Fuzzy logic; Gray scale; Median Filter; Mean Filter; Gaussian noise; Impulse
noise; Multiplicative Noise; Correction term.

1. Introduction
Digital images are used in various applications in todays life. Digital images are corrupted by noise during image acquisition or transmission process. There are different types of noises in digital images. For example, Additive white Gaussian noise (AWGN) is due to image sensors operating at low light levels, poor image acquisition or by transferring the image data in noisy communication channels. Gaussian noise is statistical noise that has its probability density function equal to that of the normal distribution, which is also known as the Gaussian distribution. In other words, the values that the noise can take on are Gaussian-distributed. Gaussian noise is properly defined as the noise with a Gaussian amplitude distribution. This says nothing of the correlation of the noise in time or of the spectral density of the noise. Labeling Gaussian noise as 'white' describes the correlation of the noise. It is necessary to use the term "white Gaussian noise" to be precise. Gaussian noise is sometimes misunderstood to be white Gaussian noise, but this is not the case. Noise is modeled as additive white Gaussian noise (AWGN), where all the image pixels deviate from their original values following the Gaussian curve. That is, for each image pixel with intensity value fij (1 i m, 1 j n for an m x n image), the corresponding pixel of the noisy image gij is given by, (1) where, each noise value n is drawn from a zero -mean Gaussian distribution as shown in fig 1.

2012, http://www.journalofcomputerscience.com - TIJCSA All Rights Reserved

10

Deepinder Kaur ,Baljit Singh, The International Journal of Computer Science & Applications (TIJCSA) ISSN 2278-1080, Vol. 1 No. 9 November 2012

Fig 1. Gaussian Distribution Curve Denoising is the pre-processing step in the Image Enhancement process. Denoising is necessary and first step to be taken before the image data is analyzed for further use. Because after introducing the noise in image, the important details and features of image are destroyed. It is necessary to apply efficient denoising technique to compensate for such data corruption. Image denoising is used to remove the noise while retaining as much as possible the important signal features. The purpose of image denoising is to estimate the original image from the noisy data.

2. Existing Methods
Generally, the Gaussian noise can easily be removed by locally averaging the pixels inside the window and replace the current pixel with this average value. Conventional linear filters such as arithmetic mean filter and Gaussian filter smooth noises effectively but blur edges. Statistical characteristics of images are of fundamental importance in removing Gaussian noise. The well known wiener filter assumes the images are second order stationary. But for most natural images the stationary assumption is not valid. The Wiener filter [2] experiences uniform filtering throughout the image, with an unacceptable blurring of fine detail across edges and inadequate filtering of noise in relatively flat areas. To overcome the problem of linear filtering, non-linear filtering techniques become popular as an alternative to preserve signal structure. Median filter [3][4]is quite popular non linear denoising filter because it provide excellent noise reduction capabilities with considerably less blurring than linear smoothening filters of same size. For impulsive noise, the median filter is one of the best. But for Gaussian noise, it is less successful. A compromise between the mean and median is trimmed mean filters. The idea behind a trimmed mean is to reject the most probable outliers-some of the very smallest and very largest values and average the rest. Alpha trimmed mean filter [5] is a one which performs the operation of both mean and median filter based on the value of . It gives better noise reduction with some smoothening capabilities. The K-nearest neighbour filter [6] also a trimmed mean filter. It handles only the values closest to the value of the center sample. The value K decides the
2012, http://www.journalofcomputerscience.com - TIJCSA All Rights Reserved

11

Deepinder Kaur ,Baljit Singh, The International Journal of Computer Science & Applications (TIJCSA) ISSN 2278-1080, Vol. 1 No. 9 November 2012

smoothening capability. In [7] Perona and Malik have proposed an anisotropic diffusion method for Gaussian noise removal. In [8] Tomasi and Manducci have proposed a bilateral filter to remove Gaussian noise with edge preservation. Recently Tamer Rabie [9] proposed a robust estimation based algorithm to remove gaussian noise. It effectively removes low to medium density Gaussian noise with edges are better preserved. In [10] R.Garnett and T.Huegerich have proposed a universal noise removal algorithm to remove Gaussian noise and other types of noises. In all these methods complexity of the algorithm is high. In this paper we proposed a Fuzzy rule based method to remove low to high density Gaussian noise with details preservation.

3. Proposed Method
Fuzzy set theory [11,12,13] has been successfully applied to pattern recognition fields. It is suitable for dealing with problems containing high levels of uncertainty, to which class pattern recognition or image processing problems usually belong. Obviously, the recovery of heavily noise-corrupted images is a task with high uncertainty levels. The general idea behind the filter is to average a pixel using other pixel values from its neighborhood, but simultaneously to take care of important image structures such as edges. The main concern of the present filter is to distinguish between local variations due to noise and due to image structure. It uses 14 fuzzy rule based convolution mask on every pixel of the image. Fuzzy membership functions used in this algorithm are not static instead it uses an adaptive approach based on the neighborhood pixels. The 3*3 mask selected for image scanning passes on pixel values to the fuzzy system input. In this mask each pixel is considered as an input image processing values between 0-255. Values of the mask are obtained by hit and trial method. Rules are determined based on pixel vicinity status. These rules have been written studying different states and special edge conditions. The considered 3*3 mask scans all image gray surfaces and pixels are examined according to predefined rules[19,20,21].

Algorithm:
Step-1: Take Noisy Image as Input image. Step2: Repeat the steps 3 to 8 for each pixel present in the input image. Step3: Consider current pixel as Center Pixel (Centre_pixel). Steps4: Clear the values Linear arrays Fuzzy[] and Possible_outcomes[]. Step5: Apply 14 Fuzzy rule based filter masks on the Centre_pixel and initialize the linear array Fuzzy[] by Fourteen different responses Ri.

2012, http://www.journalofcomputerscience.com - TIJCSA All Rights Reserved

12

Deepinder Kaur ,Baljit Singh, The International Journal of Computer Science & Applications (TIJCSA) ISSN 2278-1080, Vol. 1 No. 9 November 2012

X X X X

X X X X X X

2012, http://www.journalofcomputerscience.com - TIJCSA All Rights Reserved

13

Deepinder Kaur ,Baljit Singh, The International Journal of Computer Science & Applications (TIJCSA) ISSN 2278-1080, Vol. 1 No. 9 November 2012

X X X

Fig 2: The collection of fuzzy masks

Step6. Update Linear array Possible_outcomes[] with the values of linear array Fuzzy[] according to the following rules: a) Add The values of linear array Fuzzy[] to Possible_outcomes[] if the values lies between the range of Value(Left_pixel) and Value(Centre_pixel). Do not repeat the contents if already exists in linear array possible_outcomes[]. b) Add The values of linear array Fuzzy[] to Possible_outcomes[] if the values lies between the range of Value(Right_pixel) and Value(Centre_pixel). Do not repeat the contents if already exists in linear array possible_outcomes[]. c) Add The values of linear array Fuzzy[] to Possible_outcomes[] if the values lies between the range of Value(Up_pixel) and Value(Centre_pixel). Do not repeat the contents if already exists in linear array possible_outcomes[]. d) Add The values of linear array Fuzzy[] to Possible_outcomes[] if the values lies between the range of Value(Down_pixel) and Value(Centre_pixel). Do not repeat the contents if already exists in linear array possible_outcomes[]. Step7: Sort the contents of linear array Possible_outcomes[] in ascending order. Step8: Find the median M from the contents of Linear array Possible_outcomes[] . Step 9: Set Value[Center_pixel]=M in output image. Step 10: Exit.

2012, http://www.journalofcomputerscience.com - TIJCSA All Rights Reserved

14

Deepinder Kaur ,Baljit Singh, The International Journal of Computer Science & Applications (TIJCSA) ISSN 2278-1080, Vol. 1 No. 9 November 2012
NOISY IMAGE

APPLY 3*3 CONVOLUTION MASK IN ALL POSSIBLE WAYS

APPLY FUZZY RULES TO SELECT THE REQUIRED POSSIBLE OUTCOMES

KNOWELDGE BASE

FIND MEDIAN OF ALL POSSIBLE OUTCOMES

CORRECTION TERM REPLACE THE CURRENT PIXEL WITH MEDIAN

Output Image

Fig 3. Flowchart for proposed algorithm

4. Results And Conclusions

The proposed algorithm has been implemented on a set of different digital images. Fuzzy Logic has been used to get the denoised image of any digital image. All the simulations are done using VB.Net.

2012, http://www.journalofcomputerscience.com - TIJCSA All Rights Reserved

15

Deepinder Kaur ,Baljit Singh, The International Journal of Computer Science & Applications (TIJCSA) ISSN 2278-1080, Vol. 1 No. 9 November 2012

Noisy Image

Mean Filter

Median Filter

Proposed Algorithm

Fig4. Comparison of Filters for Figure 1.tif with noise variance 2.6

Table 1. Comparison of MSE of existing filters with proposed algorithm for 1.tif Variance(2) Mean Filter Median Filter Proposed Algorithm 1 0.0077 0.0011 0.0006 1.4 0.0043 0.0005 0.0002 1.8 0.0026 0.0003 0.0001 2.2 0.0020 0.0002 0.0001 2.6 0.0016 0.0002 0.0001

2012, http://www.journalofcomputerscience.com - TIJCSA All Rights Reserved

16

Deepinder Kaur ,Baljit Singh, The International Journal of Computer Science & Applications (TIJCSA) ISSN 2278-1080, Vol. 1 No. 9 November 2012

Table 2. Comparison of RMSE of existing filters with proposed algorithm for 1.tif Variance(2)
Mean Filter Median Filter Proposed Algorithm

1
0.0880 0.0341 0.0246

1.4
0.0657 0.0238 0.0170

1.8
0.0517 0.0182 0.0130

2.2
0.0448 0.0156 0.0113

2.6
0.0408 0.0142 0.0103

Table 3. Comparison of SNR of existing filters with proposed algorithm for 1.tif Variance(2)
Mean Filter Median Filter Proposed Algorithm

1
0.5260 0.7862 2.2095

1.4
0.3483 0.5359 1.6426

1.8
0.2558 0.3984 1.2675

2.2
0.2123 0.3322 1.0701

2.6
0.1863 0.2930 0.9427

Table 4. Comparison of PSNR of existing filters with proposed algorithm for 1.tif Variance(2)
Mean Filter Median Filter Proposed Algorithm

1
69.231 77.456 80.279

1.4
71.773 80.587 83.519

1.8
73.854 82.900 85.801

2.2
75.093 84.227 87.064

2.6
75.897 85.032 87.829

2012, http://www.journalofcomputerscience.com - TIJCSA All Rights Reserved

17

Deepinder Kaur ,Baljit Singh, The International Journal of Computer Science & Applications (TIJCSA) ISSN 2278-1080, Vol. 1 No. 9 November 2012

0.009 0.008 0.007 0.006 0.005 MSE 0.004 0.003 0.002 0.001 0 1 1.4 1.8 VARIANCE 2.2 2.6

Mean Filter Median Filter

Fuzzy Filtering

Fig 6. Comparison of MSE of existing filters with proposed algorithm for 1.tif

0.1 0.09 0.08 0.07 0.06 RMSE 0.05 0.04 0.03 0.02 0.01 0
1 1.4

Mean Filter Median Filter Fuzzy Filtering

VARIANCE

1.8

2.2

Fig 7. Comparison of RMSE of existing filters with proposed algorithm for 1.tif

2012, http://www.journalofcomputerscience.com - TIJCSA All Rights Reserved

18

Deepinder Kaur ,Baljit Singh, The International Journal of Computer Science & Applications (TIJCSA) ISSN 2278-1080, Vol. 1 No. 9 November 2012
2.5 2 1.5 SNR 1 0.5 0 1 1.4 1.8 VARIANCE 2.2 2.6 Mean Filter Median Filter Fuzzy Filtering

Fig 8. Comparison of SNR of existing filters with proposed algorithm for 1.tif
100 90 80 70 60 PSNR 50 40 30 20 10 0 1 1.4 1.8 VARIANCE 2.2 2.6

Mean Filter Median Filter Fuzzy Filtering

Fig 9. Comparison of PSNR of existing filters with proposed algorithm for 1.tif This thesis has briefly overviewed the methods for Gaussian Noise Removal so many methods have been proposed till now but the proposed algorithm has shown better results. The work presented in this thesis concludes that Fuzzy Based Approach is best method for yielding denoised images provided appropriate Fuzzy rules are chosen. Improved value of SNR, PSNR and minimized value of MSE and RMSE of proposed algorithm shows that objective performance improvement is achieved. In the proposed method all the techniques and operations provide an efficient working and the output image is enhanced according to the users requirements. Proposed Filter can clean an image completely of noise without making it blurry.

2012, http://www.journalofcomputerscience.com - TIJCSA All Rights Reserved

19

Deepinder Kaur ,Baljit Singh, The International Journal of Computer Science & Applications (TIJCSA) ISSN 2278-1080, Vol. 1 No. 9 November 2012

References
[1] A.K.Jain, Fundamentals of digital image processing, Prentice Hall, Englewood cliffs, 1989. [2] Gonzalez R.C,, Woods R.E: Digital Image Processing, Addison-Wesley, 2003 [3] P.M.Narendra. A separable median filter for image noise smoothing. IEEE Trans., PAMI3(1):20 29, 1981. [4] T. S. Huang, G. J. Yang, and G. Y. Tang. A fast two-dimensional median filtering algorithm. IEEE Trans., ASSP-27(1):1318, 1979. [5] J. B. Bednar and T. L. Watt, Alpha-trimmed means and their relationship to median filters, IEEE Trans. Acoust.,Speech, Signal Processing, vol. ASSP-32, pp. 145-153, 1984. [6] J. Astola and P. Kuosmanen, Fundamentals of Nonlinear Digital Filtering. Boca Raton, FL: CRC, 1997. [7] P. Perona and J. Malik, Scale-space and edge detection using anisotropic diffusion, IEEE Trans. Pattern Anal.Machine Intell., vol. 12, pp. 629639, 1990. [8] C. Tomasi and R. Manduchi, Bilateral filtering for gray and color images, IEEE Int. Conf. Computer Vision, pp.839846,1998. [9] T. Rabie ,Robust Estimation Approach for Blind Denoising, IEEE Transaction on Image Processing, Vol 14,No 11, pp.1755-1765 , Nov 2005. [10] Roman Garnett, T.Huegerich, A Universal Noise Removal Algorithm with an impulse Detector, IEEE Transaction on Image Processing, Vol 14,No 11, pp.1747-1754 , Nov 2005. [11] F.Russo, " A Method Based on Piecewise Linear Models for Accurate Restoration of Images Corrupted by Gaussian Noise",IEEE Trans. on Instrumentation and Measurement,Vol.55,No.6,2006. [12] F.Russo and G.Ramponi, " A noise smoother using cascaded FIRE filters", Proc. of 4th Intl. Conf. on Fuzzy Systems,Vol.1,pp. 351-358, 1995. [13] F.Russo, " A technique for image restoration based on recursive processing and error correction", Proceeding of IEEE Instrumentation and Measurement Technology Conference, Vol.3,pp.1232-1236,2000. [14] Vijaykumar V.R., Vanathi P.T., Kavagasabapathy P., Adaptive Window Based Efficient Algorithm for Removing Gaussian Noise in Gray Scale and Color Images, International Conference on Computational Intelligence and Multimedia Applications, pp. 319-323, 2007. [15] Ville D.V.D., Nachtegoel M., Naken D.V.D, Karre E.E., Philips W., Lemahier I., Noise Reduction by Fuzzy Image Filtering, IEEE Transactions on Fuzzy Systems, Vol.11, No.4, pp. 429-436, aug 2003. [16] Wilsey M., Nair M.S., Fuzzy Approach for Restoring Color Images Corrupted with Adaptive Noise, Proceedings of the World Congress on Engineering, Vol. 1, pp. 2008. [17] W. Luo, " An Efficient Detail-Preserving Appraoch for Removing Impulse Noise in Images", IEEE Signal Processing Letters, Vol. 13, NO. 7, July 2006. [18] X.Zheng and Q.Gao,"Image NOise Removal Using Perceptual Edge Features",GVIP Special Issue on Denoising, pp.15-20,2007
2012, http://www.journalofcomputerscience.com - TIJCSA All Rights Reserved

20

Deepinder Kaur ,Baljit Singh, The International Journal of Computer Science & Applications (TIJCSA) ISSN 2278-1080, Vol. 1 No. 9 November 2012

[19] A.Taguchi, H.Takahima, and F.Russo," Data Dependent Filtering using The Fuzzy Inference", Proceedings of IEEE Instrumentation and Measurement Technology Conference, Vol. 52, pp.752-756,1995. [20] D.Van De Ville, M.Nachtegael,D.Van der Weken, E.E.Kerre, W.Philips, and I.Lemahie, "Noise Reduction by Fuzzy Image Filtering", IEEE Trans. on Fuzzy Systems, Vol.11,No. 4,pp. 429-436,2003. [21] E.Abreu, M.Lightstone, S.Mitra, and K. Ara kawa, " A new efficient approach for the removal of impulse noise from highly corrupted images", IEEE Transaction on Image Processing,Vol.5,pp.1012-1025,june 1996. [22] Farbiz F, Menhaj M.B., An Extended Iterative Method for Image Enhancement Based on Fuzzy Logic, Third International conference on Knowledge-Based Intelligent Information Engineering Systems, Adelaide, Australia, pp.250-254,1999. [23] Farbiz F., Mehnaj M. B., Motamedi S. A., Hogan M. T., A New Fuzzy Logic Filter for Image Enhancement, IEEE Transaction on Systems, Man and Sybermatics- Part B. Cybermatics, Vol. 30, No. 1, pp. 110-119, 2009.

2012, http://www.journalofcomputerscience.com - TIJCSA All Rights Reserved

21