Entertainment & Media

A Novel Filter for Reduction of Random Valued Impulse Noise

Description
A Novel Filter for Reductio of Radom Valued Impulse Noise S.Gopiath 1 & C.Nivedityaa 1,,Lecturer, Departmet of ECE, 1 Muthayammal Egieerig College,Rasipuram Gaamai College of Techology, Namakkal Abstract
Published
of 5
All materials on our website are shared by users. If you have any questions about copyright issues, please report us to resolve them. We are always happy to assist you.
Related Documents
Share
Transcript
A Novel Filter for Reductio of Radom Valued Impulse Noise S.Gopiath 1 & C.Nivedityaa 1,,Lecturer, Departmet of ECE, 1 Muthayammal Egieerig College,Rasipuram Gaamai College of Techology, Namakkal Abstract A ovel method for the removal of impulse oise from digital images is itroduced. I proposed filter, threshold is computed locally from image pixels itesity values i a widow ad a detail preservig oise filter. Iitially, detectio phase classifies ay possible impulsive oise pixels. Subsequetly, the filterig phase replaces the detected oise pixels where fuzzy reasoig is employed to deal with ucertaities. Results show that proposed method provides better performace i terms of PSNR tha may other media filter variats for radom-valued impulse oise. Keywords- Radom Valued Impulse Noise, Detail preservig, PSNR, Fuzzy Reasoig, Ucertaity. I. INTRODUCTION A. Image Deoisig Digital images play a importat role both i daily life applicatios such as Satellite televisio, Magetic Resoace Imagig, Computer Tomography as well as i areas of research ad techology such as geographical iformatio systems ad astroomy.digital images are ofte corrupted by differet types of oise durig its acquisitio ad trasmissio phase. The most commo types of oise corruptig the digital images are Gaussia oise ad Impulse oise. Media filterig is a efficiet oliear techique used for impulse oise removal. Covetioal filters such as Gaussia filter used for Gaussia oise removal smooths the oise but blur the edges. To overcome these drawbacks oliear methods are proposed. Some degree of oise is always preset i ay electroic device that trasmits or receives a sigal. For televisio this sigal is the broadcast data trasmitted over cable or received at the atea; for digital cameras, the sigal is the light which hits the camera sesor. Eve though oise is uavoidable, it ca become so small relative to the sigal that appears to be oexistet. The overall oise characteristics i a image deped o may factors, icludig the type of sesor, pixel dimesios, temperature, exposure time, ad speed. So Image Deoisig is a fudametal problem i the field of image processig. The goal of image deoisig is to remove the oise while retaiig the importat sigal features. The deoisig of a atural image corrupted by Gaussia oise is a importat problem i image processig. Image deoisig still remais a challege for researchers because oise removal itroduces artifacts ad causes blurrig of the images. B. Radom Valued Impulse Noise There are differet kids of oises that will affect a image durig trasmissio through chaels, malfuctioig of camera sesors etc. The explaatio about differet kids of oises is to kow about types of oises that are affectig the images. The preset project deals with impulse oise ad a special kid of oise called radom valued impulse. The ormal values of impulse oise is 0 or 55, Where as the oise value of radom valued impulse oise ca be of ay umber betwee 0 to 55,0 or 55.This oise is itroduced ito the image by replacig the oise free pixels i the origial image by some radom values. To fid the oise value i a give image which is affected by radom valued impulse oise (RVIN), the procedure is as follows: 1) To fid the pixel values of all the pixels preset i a image. ) To icremet the cout if pixel values are same. 3) Fially the pixel value which is havig maximum cout will be treated as the oise value of the image which is affected by radom valued impulse oise. II. RELATED WORK Webi Luo [1] proposed the ew impulse oise removal techique to restore digital images corrupted by impulse oise. The algorithm is based o fuzzy impulse detectio techique, which ca remove impulse oise efficietly from highly corrupted images while preservig image details. The proposed method used to remove impulse oise i may cosumer electroics products such as digital cameras ad digital televisio (DTV) for its performace ad simplicity. Yiqiu Dog [] preseted a image statistic for detectig radom-valued impulse oise. By this statistic, most of the oisy pixels ca be modified i the corrupted images. Combiig it with a edge-preservig regularizatio, a powerful two-stage method ca be obtaied for deoisig radom-valued impulse oise, eve for oise levels as high as 60%. 7 P a g e Key Kal Vi Toh [3] preseted a ew fuzzy switchig media (FSM) filter employig fuzzy techiques i image processig. The proposed filter is able to remove saltad- pepper oise i digital images while preservig image details ad textures very well. By icorporatig fuzzy reasoig i correctig the detected oisy pixel, the low complexity FSM filter is able to outperform some well kow existig salt-ad- pepper oise fuzzy ad classical filters. Haidi Ibrahim [4] proposed a simple, yet efficiet way to remove impulse oise from digital images. This ovel method comprises two stages. The first stage is to detect the impulse oise i the image. I this stage, based o oly the itesity values, the pixels are roughly divided ito two classes, which are oise-free pixel ad oise pixel. The, the secod stage is to elimiate the impulse oise from the image. I this stage, oly the oise-pixels are processed. The oise-free pixels are copied directly to the output image. The method adaptively chages the size of the media filter based o the umber of the oise-free pixels i the eighborhood. Piar Civicioglu [5] preseted a ovel method for the suppressio of Radom-Valued Impulsive Noise from corrupted images. The proposed method is composed of a efficiet oise detector ad a pixel-restoratio operator. The oise detector has bee used to discrimiate the ucorrupted pixels from the corrupted pixels. The oise-free itesity values of the corrupted pixels have bee computed by usig Triagle-Based Liear Iterpolatio ad the values of tuig parameters of the proposed method have bee optimized with Differetial Evolutio algorithm. Key Kal Vi Toh [6] proposed a ovel two-stage oise adaptive fuzzy switchig media (NAFSM) filter for salt-ad-pepper oise detectio ad removal. Iitially, the detectio stage will utilize the histogram of the corrupted image to idetify oise pixels. These detected oise pixels will the be subjected to the secod stage of the filterig actio, while oise-free pixels are retaied ad left uprocessed. The, the NAFSM filterig mechaism employs fuzzy reasoig to hadle ucertaity preset i the extracted local iformatio as itroduced by oise. Fabrizio Russo [7] preseted a ew operator which adopts a fuzzy logic approach for the ehacemet of images corrupted by impulse oise. The proposed operator is based o two-step fuzzy reasoig, ad it is able to perform a very strog oise cacellatio while preservig image details very well. The ew fuzzy filter is favourably compared with other oliear operators i the literature. Yiqiu Dog et.al [8] proposed ew impulse detector, which is based o the differeces betwee the curret pixel ad its eighbours aliged with four mai directios. The, combie it with the weighted media filter to get a ew directioal weighted media (DWM) filter. III. OBJECTIVE OF THE PROPOSED FILTER The proposed filter operates o impulse oise desities without jeopardizig image fie details ad textures. Fast ad automated algorithm is focused. The proposed filter does ot require ay tedious tuig or time cosumig traiig of parameters as well. No priori threshold is to be give. Istead, the threshold is computed locally from image pixels itesity values i a slidig widow usig weighted statistics. More precisely, the weighted mea value ad the weighted stadard deviatio are estimated i the curret widow. The weights are the iverse of the distace betwee the weighted mea value of pixels i a give widow ad the cosidered pixel. A result is that impulse oise does ot corrupt the determiatio of these statistics from which the Threshold is derived. Noise-free pixels are relatively easy to be selected by utilizig the biary decisio. A limit for widow is set to cotai a miimum umber of pixels avoid loss of image details.i filterig mechais the proposed filter adopts fuzzy reasoig to deal with ucertaities preset i the local iformatio. These ucertaities, e.g. thi lies or pixels at edges beig mistake as oise-pixels, are caused by the oliear ature of impulse oise. The fuzzy set is processed by calculatig local iformatio to produce a suitable fuzzy membership value. IV. PROPOSED ALGORITHM I a image cotamiated by radom-valued impulse oise, the detectio of oisy pixel is more difficult i compariso with fixed valued impulse oise, as the gray value of oisy pixel may ot be substatially larger or smaller tha those of its eighbours. Due to this reaso, the covetioal media-based impulse detectio methods do ot perform well i case of radom valued impulse oise. The umerical Threshold value is defied a priori or chose after may data depedat tests. The literature shows that a optimal threshold i the sese of the mea square error ca be obtaied for most real data. However, Threshold suitable for a particular image is ot ecessarily adapted to aother oe. To overcome this proble the followig algorithm is proposed, Step 1: Read the iput image ad add Radom Valued Impulse Noise to the image. Step : Compute the weighted mea value of the widow. wm, X i j M ( (1) wm, Step 3: Weighted stadard deviatio is calculated usig the weighted mea value. 73 P a g e ( w X i j M ( ( () w Step 4: Threshold is obtaied from the above statistical parameters which is give by 1 ( i., Step 5: Noisy pixel is foud whe differece betwee cetre pixel ad weighted mea exceeds threshold. Step 6: Biary flag represets as follows: 1- Noisy pixel 0- Noise free pixel. Step 7: Compute the media value for oise free pixels. Step 8: Determie absolute differece. D( max{ d( )} (3) ) ( x) Where d( ), m) w ( 55 x ( is the cetre pixel i widow. Step 9: Compute the fuzzy membership value F ( F( 0 D( - T1 T T1 1 : D( T1 : D( T ) : T D( T1 (4) Step 10: Compute the restoratio term as follows: y( F(. M( [1 F( ]. (5) V. PERFORMANCE EVALUATION A. Peak-Sigal-To-Noise Ratio The phrase Peak Sigal to Noise Ratio, ofte abbreviated PSNR, i a egieerig term for the ratio betwee the maximum possible power of a sigal ad the power of corruptig oise that affects the fidelity of its represetatio. Because may sigal have a very wide dyamic rage, PSNR is usually expressed i terms of the logarithmic decibel scale. The PSNR is most commoly used as a measure of quality of recostructio of lossy compressio codes. The sigal i this i case is the origial data, ad the oise is the error itroduced by compressio whe comparig compressio codes it is used as a approximatio to huma perceptio of recostructio quality, therefore i some cases oe recostructio may appear to be closer to the origial tha aother, eve though it has a lower PSNR (a higher PSNR would ormally idicate that the recostructio) is of higher quality. Oe has to be extremely careful with the rage of validity of this metric; it is oly coclusively valid whe it is used to compare results from the same codec ad same cotet. B. Mea Square Error It is most easily defied via the MSE which for two M x N moochrome images I ad K where oe of the images is cosidered oisy ad approximatio of the other is defied as MSE MN PSNR M N I ( K( i0 10log 10 j0 55 MSE The maximum possible pixel value is 55. C. Mea Absolute Error Sice the PSNR by itself caot characterize the detail preservatio behavior of a filter the mea absolute error (MAE) is used to focus o the detail preservig characteristic of the filter. The MAE is defied i as: D. Results MAE MN M N I ( K( i0 j0 I order to access the performace of the proposed scheme the Lea ad Goldhill image with 51 x 51 are used as test images. This sectio summarizes all the experimetal results obtaied i adoptig the proposed method ad existig method o the radom valued impulse oise to achieve the expected result ad the performace is compared. The efficacy of the proposed method is demostrated by extesive simulatios. The experimetal results exhibit sigificat improvemet i the performace over several other methods. (A) (B) (C) (D) (E) 74 P a g e Figure 1. Results for differet methods ad the proposed method for 30% corrupted Lea image (A) origial image (B) oisy image (C) DWM (D) ASWM (E) Proposed method TABLE.PSNR COMPARISON FOR GOLD HILL IMAGE NOISE DENSITY (%) ASWM DWM PROPOSED METHOD (A) (B) VI. CONCLUSION (C) (D) (E) Figure. Results for differet methods ad the proposed method for 30% corrupted Gold hill image (A) origial image (B) oisy image (C) DWM (D) ASWM (E) Proposed method. TABLE 1.PSNR COMPARISON FOR LENA IMAGE NOISE DENSITY (%) ASWM DWM PROPOSED METHOD The proposed method does ot eed a a priori Threshold as i the case i classical Switchig Media filter to detect oisy pixels. Istead, Threshold is computed locally from image pixels grey values i a slidig widow. Fuzzy reasoig is embedded as part of its filterig mechais which permits us to exploit the effectiveess of fuzzy paradigm i hadlig imprecise local iformatio. Extesive simulatio results verify its excellet impulse detectio ad detail preservatio abilities by attaiig the highest PSNR ad lowest MAE values across a wide rage of oise desities. Thus rampat loss of image is reduced without jeopardizig image fie details. REFERENCES [1]. W. Luo, Efficiet removal of impulse oise from digital images, IEEE Tras. Cosumer Electro., vol. 5, o., pp , May 006. []. Y. Dog, R. H. Cha, ad S. Xu, A detectio statistic for radom-valued impulse oise, IEEE Tras. Image Process., vol. 16, o. 4, pp , Apr [3]. K. K. V. Toh, H. Ibrahi ad M. N. Mahyuddi, Salt-ad-Pepper Noise Detectio ad Reductio Usig Fuzzy Switchig Media Filter, IEEE Tras. Cosumer Electro., vol. 54, o. 4, pp , Nov [4]. H. Ibrahi N. S. P. Kog, ad T. F. Ng, Simple adaptive media filter for the removal of impulse oise from highly corrupted images, IEEE Tras. Cosumer Electro., vol. 54, o. 4, pp , Nov [5]. P. Civicioglu, Removal of radom-valued impulsive oise from corrupted images, IEEE Tras. Cosumer Electro., vol. 55, o. 4, pp , Nov P a g e [6]. K. K. V. Toh ad N. A. Mat Isa, Noise adaptive fuzzy switchig media filter for salt-ad-pepper oise reductio, IEEE Sigal Process. Lett. vol. 17, o. 3, pp , Mar [7]. F. Russo ad G. Rampo A fuzzy filter for images corrupted by impulse oise, IEEE Sigal Process. Lett. vol. 3, o. 6, pp , Jue [8]. Y. Dog ad S. Xu, A ew directioal weighted media filter for removal of radom-valued impulse oise, IEEE Sigal Process. Lett. vol. 14, o. 3, pp , Mar [9]. J. Zhag, A efficiet media filter based method for removig radom valued impulse oise, Digit. Sigal Process, vol. 0, o. 4, pp , July 010. [10]. Smaïl Akkoul, Roger Lédée, Remy Lecoge, ad Rachid Harba, A New Adaptive Switchig Media Filter, IEEE Sigal Processig Letters, 010, pp P a g e
Search
Related Search
We Need Your Support
Thank you for visiting our website and your interest in our free products and services. We are nonprofit website to share and download documents. To the running of this website, we need your help to support us.

Thanks to everyone for your continued support.

No, Thanks