A SURVEY ON SWITCHING MEDIAN FILTERS FOR IMPULSE NOISE REMOVAL

Transcription

1 Journal of Advanced Research in Engineering & Technology (JARET) Volume 1, Issue 1, July Dec 2013, pp , Article ID: JARET_01_01_006 Available online at IAEME Publication A SURVEY ON SWITCHING MEDIAN FILTERS FOR IMPULSE NOISE REMOVAL Sreejith L.Das Research Scholar Sathyabama University, Chennai, Alamelu Nachiappan Supervisor, Sathyabama University, Chennai, ABSTRACT Impulse noise is one of the major setbacks in digital image processing. Impulse noise can corrupt the images and also damage the fine details in the image. Various filtering algorithms are there to remove the impulse noise that is present in image during acquisition. Majority of the image denoising filters are based on the median filter. Survey on Variants of median filters like weighted median filter, adaptive switching median filter, triststate switching median filter, modified decision based unsymmentric tristate median filter are discussed in this paper. Key words: Denoising, Filtering, Weighted Median Filter, Center Weighted Median Filter, Adaptive Switching Median Filter Cite this Article: Sreejith L.Das and Alamelu Nachiappan. A Survey on Switching Median Filters for Impulse Noise Removal. Journal of Advanced Research in Engineering & Technology, 1(1), 2013, pp INTRODUCTION Filtering is an essential part of any image processing system. There are mainly two types of filter, linear and non-linear filter. When the noise is non-additive, linear filtering techniques fall short and are not effective in removing impulse noise. Whereas the non-linear filter algorithm are often adopted for the removal of salt and pepper noise. Digital images are normally corrupted by many types of noise [1-2]. The major cause of impulse noise is due to faulty switching elements or poor lighting conditions during the signal acquisition. The non-liner filtering techniques are preferred for removing the impulse noise, because they are having good edge and image detail preservation properties. The median filters are the most commonly used filter for the removal of salt and pepper noise. They modify both the noise as well as the noise free pixels resulting in blurred and distorted features. Modified forms of median filter were introduced to overcome 58 editor@iaeme.com

2 A Survey on Switching Median Filters For Impulse Noise Removal this problem. Various filtering algorithm are used for the removal of impulse noise. This survey details about various filtering techniques to remove the impulse noise. 2. VARIOUS FILTERING TECHNIQUES This paper deals with the study of various median filtering techniques to remove the impulse noise from the digital images A. Standard Median Filter The standard median filter (SMF) has been established as consistent method to remove the salt and pepper noise without damaging the edge details [3-5]. The impulse noise is removed by changing the intensity value of the center pixel of the window with the calculated median intensity value of the pixel within the window. SMF removes thin lines and blur image details even at low noise densities. SMF is unable to differentiate the corrupted pixels with uncorrupted ones. It does not work well with the large filter size. The major drawback of the standard median filter is that the filter is effective only at low noise density. When the noise density is high, it will not preserve the edge details of the original image. The disadvantage is that many noise free pixels get replaced in the image. B. Weighted Median Filter Weighted median filter is one of the extension or subdivision of the standard median filter [6-7]. The operations involved in the weighted median filter are same as the standard median filter. The only difference is that, the weighted median filter uses a weighted vector for the calculation of the median value. These weights correspond to the number of sample duplication. S(i,j)=median(k,l) Ԑ Ẉm,n{D(i+k,j+l)} The median value of an image is calculated using the above equation. A sliding window of size mxn pixels centered with coordinates (i,j) is taken and the weight is assigned such that the weight decreases when it is away from the center of the filtering window. This way of processing will give more importance to the center pixel and thus the noise is suppressed. In weighted median filter, the image details are preserved to a large extend. The preservation of the image detail depends on the weighing coefficient. In practical it is difficult to find the weighing coefficient for this filter. The median value taken from the sorted list of samples manipulated with the weight gives the weighted median output. It requires large computational time. C. Center Weighted Median Filter Center weighted median filter (CWMF) is known as detail preserving filter [8-9]. It is a subclass of the weighted median filter. The filter gives more weight to the center pixel in the window. CWM filter is mainly based on statistical and deterministic properties. In this filter, the center pixel is assigned a larger value of weight, w (i, j) = (2k+1) and all other pixel values are assigned one w(i-1,j-1) to w(i+1, j+1) = 1. W= Center weight median filter is specified by two parameter: The size of the window and the center weights. The behavior of the Center weighted median filter (CWMF) can be easily adjusted by changing the center weight of the window. It is easy to 59 editor@iaeme.com

3 Sreejith L.Das and Alamelu Nachiappan implement. It can preserve fine details and suppress noise particularly high density noise. The disadvantage is that it misses many impulses. Many good pixel are also modified results in the blurring of images. D. Recursive Weighted Median Filter The sample weights are assigned according to the changes by the low pass filter. RWMF detects and remove the impulse in the image [10-11]. Y(n)= In RWMF the output is formed from the previous output and the input. The weight calculation for the recursive weight median filter is performed by many techniques like threshold decomposition technique, optimization technique. In optimization technique MAE and MSE are used for calculating the weights, MAE technique is used for real weight calculation and complex calculation. For each window, the input sample which is closer to the output of the first filtering operation can be weighted more. The first iteration output is taken as the reference signal and computing the new weights and then comparing the new reference signal to the original signal and computes the output. This is repeated until the required number of iterations has been attained. Reference signal calculation is done only for the first time. RWM filter produces better results for high density noises. RWM filter using Lin s algorithm which produces less efficient output and the algorithm has high complexity. E. Adaptive Switching Median Filter Adaptive switching median filter (ASMF) is proposed to remove the salt and pepper noise from the corrupted images [12-14]. The algorithm used for the ASMF is they are combining the advantages of the PSMF and the adaptive median filter. The salt and pepper noise is considered. The idea of this filter is based on the impulse noise detection. The noisy pixel replacement is easy once the noise and their position are detected in the image by the best estimate of the good pixel. The binary value 0 and 1 are used to indicate whether the pixel is good or bad. If the binary value indicates 0 means the pixel is good and 1 means that has been found to be an impulse noise. Only the pixel which is affected by noise will be processed. The advantage of this filter is that they can remove almost the entire noise pixel. It is used for edge detection and object recognition. F. New Tristate Switching Median Filter Switching median filter and the decision based median filter are suitably combined together called new tristate switching median filter (NTSMF)[15-16]. This filter proposes to suppress noise and also to preserve the useful information in the image. The impulse detector is used to detect the impulse noise from sources. This is based on the switching logic. Once the noise is detected, the output of the decision based filter is taken and then compare with the center pixel value within the window in order to make decision. To improve the noise detection, the absolute difference between original pixel value from noisy image and center pixel value from the filter image is compared with predetermined threshold values. Among these, a particular value which is the optimum value for quantitative and qualitative measure is taken and given for the 60 editor@iaeme.com

4 A Survey on Switching Median Filters For Impulse Noise Removal filtering process. The filtering technique has better performance in terms of both quantitative and qualitative measure. G. Modified Decision Based Unsymmetric Trimmed Median Filter Modified decision based unsymmetric trimmed median filter (MDBUTMF) can restore images that are highly corrupted by salt and pepper noise [17-19]. Because at high density DBUTMF does not give better performance for the removal of noise present in image. The presence of noise is checked in each and every pixel of the image. If the selected window contain noise and the neighboring pixel also contains noise then the median value again will become noisy. In such case, the mean value of the window is taken and then processes it. In other case 1-d array of the selected image is taken and the salt and pepper noise is eliminated and then the median of the array value is calculated. The median value then replaces the processing pixel. If the window is noise free then it does not require further processing. It has better performance than the other filter. 3. RESULTS AND DISCUSSION This paper surveys different median filtering technique like standard median filter(smf), Weighted median filter(wmf),center weighted median filter(cwmf), Adaptive switching median filter(asmf), Recursive weighted median filter (RWMF),, New tristate switching median filter(ntsmf), and Modified decision based unsymmetric trimmed median filter(mdbutmf). Standard median filter is used for the removal of impulse noise. The disadvantage of the SMF is that many noise free pixels are removed in the image. Weighted median filter performs the same operation as SMF. The advantage is that fine details are preserved. The disadvantage is that it requires high complex time. Center weight median filter is mainly based on statistical and deterministic properties. Advantage of CWMF is it can suppress heavy noise and the disadvantage is that it misses many impulses, many good pixel are also modified. RWMF detects and remove the impulse in the image. The disadvantage is that RWM filter using Lin s algorithm produces less efficient output and the algorithm has high complexity. Adaptive switching median filter (ASMF) is proposed to remove the salt and pepper noise from the corrupted images. It is used for edge detection and object recognition. The disadvantage is that time required for computation is more. NTSM filter proposes to suppress noise and also to preserve the useful information in the image. NTSMF does not give better performance. MDBUTM filter can restore images that are highly corrupted by salt and pepper noise. It has been found from the survey that MDBUTM filter is better than the other filters which are discussed. The filter performance is measured using the peak signal to noise ratio. It is also used to process the color images that are corrupted by salt and pepper noise. This is effective for the removal of noise at high densities Table 1 tabulates a subjective comparison of the restored test image, cameraman, produced by SMF, WMF, CWMF, RWMF, ASMF, NTSMF, and MDBUTMF respectively editor@iaeme.com

5 Sreejith L.Das and Alamelu Nachiappan TABLE 1 PSNR OF FSMF, PSMF, TSMF, ASMF AND HSF AT DIFFERENT NOISE LEVELS Noise Ratio 10% 20% 30% 40% 50% 60% 70% 80% SMF WMF CWMF RWMF ASMF NTSMF MDBUTMF CONCLUSION In this paper different type of median filter used for the removal of impulse noise have been reviewed. The performance of MDBUTMF is found better than that of other filter because they can denoise even at the higher density and the PSNR value of the MDBUTMF is higher. So it is observed that the MDBUTMF is performing better when compared with other. REFERENCES [1] R.C Gonzalez, R.E.Woods Digital Image Processing 3 rd edition, Prentic Hall Publication. [2] S.E.Umbaugh, Computer Vision and Image Processing, Prentice-Hall Engelwoods Clifts,NJ,USA,1998 [3] N.C.Gallagher, Jr.G.L.Wise, A theoretical analysis of the properties of median filters, IEEE Trans accoust, Speech and signal,vol.assp.29,pp ,dec.1981 [4] R.K.Yang, L.Yin, M.Gabbouj, J.Astola, and Y.Neuro, Optimal Weighted Median Filtering Under Structural Constraints, IEEE Transcation on Signal Processing, 1995,vol 43,no.3,pp [5] T.S.Hung,G.J.Yang and G.Y.Tang A fast two dimensional median filtering algorithm IEEE Transaction on acoustics, speech and signal processing,1979,vol.27, no. 1,pp13-18 [6] Xuming Zhang and Youlun Xiong, Impulse Noise Removal Using Directional Difference Based Noise Detector and Adaptive Weighted Mean Filter, IEEE Signal Letters,vol.16,No,.4 April 2009 [7] Y.Q.Dong and S.F.Xu A new directional weighted median filter for removal of random valued impulse noise IEEE signal processing letters, 2007,vol.14,no.3.pp [8] S.J.Ko and Y.H.Lee,1991, Center Weighted Median Filter, IEEE Transaction, pp [9] T.Sun, Center weighted median filters: some properties and their applications in image processing, Signal processing, 1994, vol.35, no.3. pp [10] V.R.Vijay Kumar, S.Manikandan, P.T.Vanathi, P.Kanagasabapathy, D.Ebeneer, Adaptive Window Length Recursive Weighted Median Filter for Removing Impulse Noise in Images with Details Preservation, ECTI Transcation Electrical Eng, Electronics, and Communication vol, No.1, Feb editor@iaeme.com

6 A Survey on Switching Median Filters For Impulse Noise Removal [11] G.R.Arce and J.L.Paredes Recursive weighted median filters admitting negative weighted and their optimizations. IEEE Transcation signal processing 2000.vol.48,no.3,pp [12] Vladimiar V.Khryaschev, Andrey L.Prioror, Illya.V.Apalkov, Pavel S.Zvonarev, Impulse Denoising Using Adaptive Switching Median Filter, IEEE Transcation. [13] H,.Ibrahim Adaptive switching median filter utilizing quantized window size to remove impulse noise from digital images. Asian transaction on fundamental of electronics communication and multimedia,2012,vol.2,no.1,pp 1-6 [14] H.Hwang and R.A.Hadded, Adaptive median filter: New algorithm and results, IEEE Trans. Image process.vol.4, no.4, pp , April [15] S.Athi Narayanan, G.Arumugam, Prof.Kamal Bijlani Trimmed Median Filter For Salt and Pepper Noise Removal, International Journal Of Emerging Trends and Technology in Computer Science, Vol 3,Issue 1,Jan-Feb [16] R.Pushpavalli, G.Sivaradje, A New Tristate Switching Median Filtering Technique For Image Enhancement, International Journal and Advance Research Engineering and Technology, Vol 3,Issue 1,Jan-June 2012 pp [17] S.Esakkirajan, T.Veerakumar, Adabalan Subramanyam and C.H.Prem chand Removal of High Density Salt and Pepper Noise Through Modified Decision Based Unsymmetric Trimmed Median Filter IEEE Signal Processing Letters, Vol 18,No.5, May 2011 [18] S.Gopi Krishna, T.Sreenivasulu Reddy, G.K.Ranjini, Removal of High Density Salt and Pepper Noise Through Modified Decision Based Unsymmetric Trimmed Median Filter, International Journal of Engineer and Research and Application, Vol 2,Issue 1,Jan 2012 pp [19] Dodda shekhar, Rangu Srikanth, Removal of high density salt and pepper noise in noisy images using decision based unsymmentric trimmed median filter International Journal of Computer Trends and Technology.vol.2.Issue pp editor@iaeme.com