Salt And Pepper Noise Detection And Removal By Modified .

Medida.Amulya Bhanu et al., International Journal of Advanced Trends in Computer Science and Engineering, 1 (3), July-August, 93-97ISSN No. 2278-3091Volume 1, No.3, July – August 2012International Journal of Advanced Trends in Computer Science and EngineeringAvailable Online at http://warse.org/pdfs/ijatcse03132012.pdfSalt and Pepper Noise Detection and removal by Modified Decision basedUnsymmetrical Trimmed Median Filter for Image Restoration11&2Medida.Amulya Bhanu, 2Gopichand Nelapati, 3Dr.Rajeyyagari SivaramNalanda Institute of Engg. & Tech. Guntur, 3Amara Institute of Engg. & Tech.3dr.r.sivaram@gmail.comnoise level is over 50% the edge details of the original imagewill not be preserved by standard median filter.ABSTRACTIn this paper, six different image filtering algorithms arecompared based on their ability to reconstruct noise affectedimages. The purpose of these algorithms is to remove noisefrom a signal that might occur through the transmission of animage. A new algorithm, the Spatial Median Filter, isintroduced and compared with current image smoothingtechniques. Experimental results demonstrate that the proposedalgorithm is comparable to these techniques. This proposedalgorithm shows better results than the Standard Median Filter(MF), Decision Based Algorithm (DBA), Modified DecisionBased Algorithm (MDBA), and Progressive Switched MedianFilter (PSMF). The proposed algorithm is tested againstdifferent grayscale and color images and it gives better PeakSignal-to-Noise Ratio (PSNR) and Image Enhancement Factor(IEF).Keywords: Median Filter, Salt andUnsymmetrical Trimmed Median Filter.PepperNoise,1. INTRODUCTIONImpulse noise in images is present due to bit errors intransmission or introduced during the signal acquisition stage.There are two types of impulse noise, they are salt and peppernoise and random valued noise. Salt and pepper noise cancorrupt the images where the corrupted pixel takes eithermaximum or minimum gray level. Several nonlinear filtershave been proposed for restoration of images contaminated bysalt and pepper noise. Among these standard median filter hasbeen established as reliable method to remove the salt andpepper noise without damaging the edge details. However, themajor drawback of standard Median Filter (MF) is that thefilter is effective only at low noise densities [1]. When theAdaptive Median Filter (AMF) [2] perform well at low noisedensities. But at high noise densities the window size has to beincreased which may lead to blurring the image. In switchingmedian filter [3], [4] the decision is based on a pre-definedthreshold value. The major drawback of this method is thatdefining a robust decision is difficult. Also these filters willnot take into account the local features as a result of whichdetails and edges may not be recovered satisfactorily,especially when the noise level is high.To overcome the above drawback, Decision Based Algorithm(DBA) is proposed [5]. In this, image is denoised by using a 33 window. If the processing pixel value is 0 or 255 it isprocessed or else it is left unchanged. At high noise density themedian value will be 0 or 255 which is noisy. In such case,neighboring pixel is used for replacement. This repeatedreplacement of neighboring pixel produces streaking effect [6].In order to avoid this drawback, Decision Based UnsymmetricTrimmed Median Filter (DBUTMF) is proposed [7]. At highnoise densities, if the selected window contains all 0’s or 255’sor both then, trimmed median value cannot be obtained. Sothis algorithm does not give better results at very high noisedensity that is at 80% to 90%. The proposed ModifiedDecision Based Unsymmetric Trimmed Median Filter(MDBUTMF) algorithm removes this drawback at high noisedensity and gives better Peak Signal-to-Noise Ratio (PSNR)and Image Enhancement Factor (IEF) values than the existingalgorithm.The rest of the paper is structured as follows. A briefintroduction of unsymmertric trimmed median filter is given inSection 2. Section 3 describes about the proposed algorithmand different cases of proposed algorithm. The detailed93@ 2012, IJATCSE All Rights Reserved

Medida.Amulya Bhanu et al., International Journal of Advanced Trends in Computer Science and Engineering, 1 (3), July-August, 93-97description of the proposed algorithm with an example ispresented in Section 4. Simulation results with differentimages are presented in Section 5. Finally conclusions aredrawn in Section 6.have improved the spatial median by deriving a fasterestimation formula [6]. The spatial depth between a point andset of points is defined by,2. SPATIAL MEDIAN FILTERWhen transferring an image,sometimes transmissionproblems cause a signal to spike, resulting in one of the threepoint scalars transmitting an incorrect value. This type oftransmission error is called “salt and pepper” noise due to thebright and dark spots that appear on the image as a result f thenoise. The ratio of incorrectly transmitted points to the totalnumber of points is referred to as the noise composition of theimage. The goal of a noise removal filter is to take a corruptedimage as input and produce an estimation of the original withno foreknowledge of the noise composition of the image.The following is the basic algorithm for determining theSpatial Median of a set of points, x1, .,xN: Let r1, r2, ., rNrepresent x1,x2, .,xN in rank order such thatIn images containing noise, there are two challenges. The firstchallenge is determining noisy points. The second challenge isto determine how to adjust these points. In the VMF, a point inthe signal is compared with the points surrounding it asdefined by a filter mask. Each point in the mask filter is treatedas a vector representing a point in a three-dimensional space.Among these points, the summed vector distance from eachpoint to every other point within the filter is computed. Thepoint in the signal with the smallest vector distance amongthese points is the minimum vector median.The point in spacethat has the smallest distance to every other point is consideredto be the best representative of the set. The original VMFapproach does not consider if the current point is original dataor not.and let rc represent the center pixel under the mask. Then,If a point has a small summed vector distance, yet is not theminimum vector median, it is replaced anyway. The advantageof replacing every point achieves a uniform smoothing acrossthe image. The disadvantage to replacing every point is thatoriginal data is sometimes overwritten. A good smoothingfilter should simplify the image while retaining most of theoriginal image shape and retain the edges. A benefit of asmoothed image is a better size ratio when the image needs tobe compressed. The Spatial Median Filter (SMF) is a newnoise removal filter. The SMF and the VMF follow a similaralgorithm and it will be shown that they produce comparableresults. To improve the quality of the results of the SMF, anew parameter will be introduced and experimental datademonstrate the amount of improvement.The idea behind a trimmed filter is to reject the noisy pixelfrom the selected 3 3window.Alpha Trimmed Mean Filtering(ATMF) is a symmetrical filter where the trimming issymmetric at either end. In this procedure, even theuncorrupted pixels are also trimmed. This leads to loss ofimage details and blurring of the image. In order to overcomethis drawback, an Unsymmetric Trimmed Median Filter(UTMF) is proposed. In this UTMF, the selected 3 3 windowelements are arranged in either increasing or decreasing order.Then the pixel values 0’s and 255’s in the image (i.e., the pixelvalues responsible for the salt and pepper noise) are removedfrom the image. Then the median value of the remaining pixelsis taken. This median value is used to replace the noisy pixel.This filter is called trimmed median filter because the pixelvalues 0’s and 255’s are removed from the selected window.This procedure removes noise in better way than the ATMF.The SMF is a uniform smoothing algorithm with the purposeof removing noise and fine points of image data whilemaintaining edges around larger shapes. The SMF is based onthe spatial median quintile function developed by P. Chaudhuriin 1996, which is a L1 norm metric that measures thedifference between two vectors [4]. R. Serfling noticed that aspatial depth could be derived by taking an invariant of thespatial median [5]. The Serfling paper first gave the notion thatany two vectors of a set could be compared based on their“centrality” using the Spatial Median. Y. Vardi and C. ZhangThe SMF is an unbiased smoothing algorithm and will replaceevery point that is not the maximum spatial depth among itsset of mask neighbors. The Modified Spatial Median Filterattempts to address these concerns.3. UNSYMMETRIC TRIMMED MEDIAN FILTER4. PROPOSED ALGORITHMThe proposed Modified Decision Based UnsymmetricalTrimmed Median Filter (MDBUTMF) algorithm processes theCorrupted images by first detecting the impulse noise. The94@ 2012, IJATCSE All Rights Reserved

Medida.Amulya Bhanu et al., International Journal of Advanced Trends in Computer Science and Engineering, 1 (3), July-August, 93-97processing pixel is checked whether it is noisy or noisy free.That is, if the processing pixel lies between maximum andminimum gray level values then it is noise free pixel, it is leftunchanged.If the processing pixel takes the maximum or minimum graylevel then it is noisy pixel which is processed by MDBUTMF.The steps of the MDBUTMF Each and every pixel of theimage is checked for the presence of salt and pepper noise.Different cases are illustrated in this Section. If the processingpixel is noisy and all other pixel values are either 0’s or 255’sis illustrated in Case i). are elucidated as follows.If the processing pixel is noisy pixel that is 0 or 255 isillustrated in Case ii). If the processing pixel is not noisy pixeland its value lies between 0 and 255 is illustrated in Case iii).Case i): If the selected window contains salt/pepper noise asprocessing pixel (i.e., 255/0 pixel value) and neighboring pixelvalues contains all pixels that adds salt and pepper noise to theimage:5. SIMULATION RESULTSThe performance of the proposed algorithm is tested withdifferent grayscale and color images. The noise density(intensity) is varied from 10% to 90%. Denoisingperformances are quantitatively measured by the PSNR andIEF as defined in (1) and (3), respectively:95@ 2012, IJATCSE All Rights Reserved