A

Survey on Various Median Filtering Techniques for

Removal

of Impulse Noise from Digital Image

ABSTRACT

One of the trending fields of image

processing is elimination of noise from a contaminated image. Image sequence

are often corrupted by noise, e.g. due to bad reception of signals over channel

through which they are transmitting. Some noise sources are located in sensor

of a camera and become active during image acquisition under bad lighting

conditions. Other noise sources are due to transmission over analogue channels

and thus need to use analog to digital converter which also leads to noise

contamination. In the most cases, it is very necessary to eliminate impulse noise

from image data. Median based filters are commonly used for the removal of such

impulse noise. However, there are different kind of median filters such as

weighted median filter, recursive median filter, iterative median filter,

directional median filter, adaptive median filter and switching median filter. Therefore,

this paper will survey different median filtering techniques for removal noise

from image.

Keywords—Adaptive

median filter, directional median filter, impulse noise, iterative median

filter, median filter, recursive median filter, switching median filter, weighted

median filter, threshold filter.

INTRODUCTION

Images

are vital and necessary for communication. Numerous images are produced,

translated and edited every day. However, as the variety of imaging devices,

the quality of images is unstable. Some information may be damaged during the

transition process between different devices. Thus, to remove noise and recover

meaningful information from noisy images is a vital problem.

Noise is the most annoying problem in

processing. It introduces random variations into image that fluctuate the

original values to some different values. One way to get rid of this problem is

the development of such a robust algorithm that can perform the processing

tasks in presence of noise. The other way is to design a filtration process to

eliminate the noise from images while preserving its features, edges and

details. The filters can be divided in two types: linear filter and non-linear

filter. Linear filters are like average filter or called averaging low pass

filter. But linear filter tends to blur edges and other details of image, which

may reduce the accuracy of output. On the other hand non-linear type filter

like median filter has better results than linear filter because median filter

remove the impulse noise without edge blurring. A median filter is an example

of a non-linear filter and it is very good at preserving image detail.

Generally different filters are used for eliminating different noises like Mean

filter is used to remove the impulse noise, median filtering is used to

eliminate salt and pepper noise and so on.

The following three steps are used to

calculate median: 1. Process each pixel in the image 2. Select the size of

filtering window and sort the all pixels of window in order based upon their intensities.3.

Replace the central pixel with the median value of sorted pixels.

The performance of median filter also

depends on the size of window of filter. Smaller window preserves the details

but it will cause the reduction in noise suppression. Larger window has great

noise reduction capability but image details (edges, corners, fine lines)

preservation is limited. With the improvement in the standard median filters,

there were so many filters has designed like weighted median filter, centre

weighted median filter ,adaptive median filter, rank order median filter and

many other improved filters. Different median filters uses different sorting

algorithm like merge sort, quick sort, and heap sort to sort the elements of

window.

LITERATURE SURVEY

In

3,Al-Araji introduced an auto selection method for reduction of impulse noise

in wireless communication systems. The selection was made on the basis of estimation of rate of impulse

noise present in the system. The impulse noise is detected by comparing it with

a fixed reference. Their system used subtraction-gating for low occurrence rate

and directly sends the incoming signal to the output. If impulse is absent. The

simulation results for QAM and FSK and real time implementation on an FPEG were

presented and the bit error rate was estimated.

In

4, Chen put forward the idea of an adaptive pixel correlation filter(ACPF) to

eliminate impulse noise, with an adaptive threshold designed based on the correlations between a

pixel and its neighbour. The filter was designed with a novel adaptive working

window and weighted function for impulse noise detection and preserving

noise-free pixels based on the fact that both horizontal and vertical

correlations for a pixel are more significant than other orientations.

In

5, Chen proposed an adaptive working window to remove impulse noise. There was

a correlation between the neighbouring pixels. The pixel correlation was used

to determine the intensity of impulse noise . They introduced a simple rule for

impulse noise reduction for various median-based filter with adaptive working

window. The switching median filter and multi-state median filter are designed

with this simple rule and the simulations showed that this method enhances the reliability.

In

6, Divyajyothi proposed a new integrated fuzzy filter for the reduction of

additive noise and impulse noise from digital color images. They used an

impulse noise detector in the filter to detect the presence of impulse. The

detector divides the set of pixels into affected points and clean points. A

selection filter module was employed to select the appropriate filter to match

the type of noise. The output of integrated filter contains the enhanced image

after noise removal. Their method combined the advantage of both additive noise

and impulse noise filters. Their results showed that the approach effectively

removed the integrated noise even at high noise levels.

In

7,Rezvanian presented an efficient method to reduce impulse noise in two

passes. During first pass impulse noise detection using ANFIS, and at the

second pass the impulse noise estimation, that corrupted noise pixel replaced

with new value based on neural networks. This method was experimented on

popular greyscale test images and compared with other methods using subjective

and objective measures. Their results showed that the method works efficiently

in reducing impulse noise.

In

8, Tao Chan proposed a tristate median filter for image denoising .This

preserves image details while effectively suppressing impulse noise. This

proposed filter outperforms other median filter by balancing the tradeoffs’

between noise reduction and detail preservation. To achieve better results, a

camera calibrations procedure may be placed before our systems. For different

values of noise density, optimum threshold range for yielding smallest MSE

values and good visual quality can be obtained through similar simulation

experiments.

In

9, Zhou Wang and David zhang proposed a new median based filter i.e. progressive

switching median filter to restore images corrupted by salt and pepper noise. This

algorithm consists of two main concepts: switching scheme based filter and progressive

method impulse detection and nose filtering procedures are progressively through

several iterations. The experiment results shows that the proposed algorithm is

better than the existing median based

filters. This method is particularly useful for images, having corruption ranging

from 5% to 70%. The MSE is also effective than other median based methods, even

when noise ratios are high. It eliminates

most of the noisy pixels along with preserving image details .

In

10 Tao Chen and Hong Ren vu introduced a novel adaptive median-filter that

consists the switching scheme based on the impulse detection mechanism. Centre

weighted median filters are used for adaptive

impulse detection. The extensive simulations shows that the proposed filter

consistently works well in suppressing both fixed and random valued impulses with

different noise ratios. While, still possessing a computational structure.

In

11, Senk and Trpovski proposed a robust estimator of the variance, MAD, is

modified and used to efficiently separate noisy pixels from the image details

and therefore has no sensitivity to image contents. The complexity of proposed

algorithm is equivalent to that of the median filter. The pixel wise MAD concept

is straight forward, low in complexity and achieves high filtering performance.

In

12, Raymond Chan,Chen Hu and Mila Nikolova proposed a two stage iterative

method for removing random valued impulse noise. In the first phase, an

adaptive centre weighted median filter(ACWM) is used to identify corrupted pixels . In the

latter phase, these noise candidates are restored using a detail preserving

regularization method which allows edges and noise free pixels to be preserved.

These two phases are applied alternatively. This simulation results indicate

that the proposed method is significantly better than those using just non

linear filters or regulization only. This can be done very fast. The timing can

be enhanced by better implementations of minimization routines for solving the

robust statistics with applications to early vision.

In

13, Hancheng Yu,Li Zhao and Haixian Wang proposed an efficent algorithm

method uses a statistics of rank ordered relative differences to identify

pixels which are likely to be corrupted .It consists of two methods 1.RORD

impulse detector into many existing techniques, allowing them to detect and

properly handle impulse like pixels. 2. A sample weighted mean filter (SWMF) by

using the RORD detector and the reference image to suppress impulse noise, while

preserving image details. Although our algorithm is applied iteratively much

faster, especially when the noise is high, the sample weighted filter offers

good filtering performance while its implementation complexity is lower than

others.

In

14, Francisco Estard, David Fleet and Allan Jepson proposed a probabilistic

algorithm for image noise removal.ie., Stochastic Image Denoising .These

proposed algorithm for image denoising based on simulated random walks on image

space and also very simple. Random walk produce stable estimates even for few

trials and the overall behaviour of the random walks approximate that of more

computationally expensive blur kernels. Stochastic denoising will become a

useful tool for noise removal.

In

15, Jain-Feng Cai,Raymond H.Chan and Mila Nikolova(2010) proposed a fast two –

phase image de blurring under impulse noise. This proposed algorithm consists of

a two- phase approach to restore image corrupted by blur and impulse noise,

which is much simpler. In the first phases, accurate detection of a location of

outlier using a median-type filter and second phase edge -preserving

restoration that de-blur using only those data samples that are note noise

candidates. The PSNR of the restoration by our method is about 1 to 3 dB higher

than that by the variation method. Even for blurred images corrupted by 55%

random valued noise, proposed method can give a very good restoration result.

Comparing the two –phase methods. Image De-blurring under impulse noise with

the two- phase methods for de-blurring images corrupted by impulse Gaussian

noise. The proposed method produces more computationally efficient and takes

only about 1/8 CPU.

In

16,P.Etrahanias and A.N.Venetsanopoulos (1993) proposed a Vector Directional

Filter-A New Class of Multichannel Image Processing Filters. These proposed

filters separate the processing of vector-valued signals into directional

processing and magnitude processing. This provide a link between single-channel

image processing, where only magnitude processing is essentially performed, and

multichannel image processing where both the direction and the magnitude of the

image vectors play an important role in the resulting image. VDF perform at

least as good and in most cases and also better chrominance estimate than VMF

and this justifies their employment in colour image processing. VDF can achieve

very good filtering results for various noise source models.

In

17, Yiqiu Dong, Raymond H, Chan and Shufang Xu (2007) proposed a Detection

Statistic for Random -Valued Impulse Noise. This proposed technique for

detecting random –valued impulse noise based on image statistic. By this

statistic, we can identify most of the noisy pixels in the corrupted images.

Combining it with an edge preserving regularization, we obtain a powerful two

-stage method for de-noising random valued impulse noise even for noise level

as high as 60%. These propose a new local image statistic ROLD, by which we can

identify more noisy pixels with less false-hits. We combine kit with the edge preserving

regularization in the two- stage iterative method.

In

18 , Yiqiu Dong and Shufang Xu proposed a New Directional weighted median

filter for removal of RVIN. They

proposed a new impulse detector .based on difference between current pixels and

its neighbours aligned with four main directions. Then we combine it with the

weighted median filter to get a new directional weighted median filter. This

proposed method suppresses noise level and preserve image details too, including

thin lines. In addition it can be

extended to restore color images corrupted by impulse noises.

In

19, Roman Garnett, timothy Huegerich and charles chui introduced a universal

noise removal algorithm, which consists of an impulse detector. This proposes a

local image statistics for finding corrupted pixels. The simulation result is

capable of reducing both Gaussian and impulse noise .This method is extended to

remove any mix of Gaussian and impulse noise.

In

20, vector directional filters (VDF) for multichannel image processing were introduced.

VDF separates the processing of vector valued signal into directional

processing and magnitude processing. This provides a link between single

channel image processing, where only magnitude processing is essentially

performed and multichannel image processing where both the direction and magnitude

of the image vectors play a vital role in the processed image. The experimental

result shows that in case of color images, VDF achieved good results for various noise source models.

In

21, shih-Chang Hsia proposed an efficient noise removal algorithm using an

adaptive digital image processing approach. Simulations have demonstrated that

the new adaptive algorithm could effectively reduce noise impulse even in

corrupted images. In order to achieve real time implementations, a cost

effective architecture is proposed using parallel structure and pipelined

processing. The proposed processor can achieve throughput rate of 45M pixels/s

using only 4k gates and two line buffers. Unlike median filtering chips, this

processor provides better filtering quality and its circuit is much less

complex.

In

22, Deepa Et Al proposed an efficient low cost VLSI architecture for the edge

preserving impulse noise removal technique. The architecture comprises of two

line buffers, register banks, impulse noise detector, edge oriented noise

filter and impulse arbiter. The proposed algorithm involves only fixed size

window instead of variable window size. The storage space required is a two

line buffer. Both, reduces storage requirement as well as computation

complexity. The implemented edge preserving algorithm results in better visual

quality and pipelined architecture results in better visual quality for

denoised image.

CONCLUSION

This

paper surveys different common median filtering techniques. Each technique has

its own advantages, and disadvantages. From literature, we found that most of

the recent median filtering based methods employ two or more than two of these

frame work in order to obtain an improved impulse noise reduction and enhancing

picture quality.

REFERENCE

1 Vikas Gupta, Madhu Shandilya, Image

De-noising by Dual Threshold Median Filtering for Random Valued Impulse Noise .

2 Youlian Zhu, Cheng Huang,An Improved Median Filtering Algorithm for Image Noise

Reduction

3 Al-Araji,S.R,M.A,AL-Qutayri,K.Belhaj

and N. Al-Shwawreh 2007. Impulsive noise reduction techniques based on rate of

occurrence estimation. 9th

International Symposium on Signal Processing and Its

Applications,1-4pp,2007.

4 Chen,T.H.,C.Y. Chen , T.Y. Chen

2006. An impulse Noise Reduction Method by Adaptive Pixel-Correlation. First

International Conference on Innovative Computing, Information and Conrol

(ICICIC’06)1,257-260.

5 T.H. Chen,S.F. Huang, C. Y.Chen and

Z.H. Lin:IIH-MSP(2006)26057.

6 D.Divya Jothi, P.Geetha and S. Anna

Durai:IEEE SSD(2008) 1.

7 Rezvanian , A.K. Faez and F.

Mahmoudi 2008. A two – pass method to impulse noise reduction from digital

images based on neural networks . International Conference on Electrical and

Computer Engineeribg(ICECE 2008),400-405.

8 T. Chen , K.K. Ma and L.H. Chen ,

Tri-state median for image denoising ,IEEE Trans, Image Process,

vol.8,no.12,pp.1834 1838,Dec .1999.

9

Z.Wang and D.Zhang,Progressive switching median filter for the removal

of impulse noise from highly corrupted images,IEEETrans,Circuits Syst. II,

vol.46, no.1,pp.78 80,1999.

10 T.Chen and H.R.Wu. Space variant median

filters for the retoration of impulse noise corrupted images ,IEEE Trans .

Circuit Syst.II ,vol.46, no.8,pp.784 789, Aug.2001.

11 T.Chen and H.R.Wu. Adaptive

impulsive detection using center weighted median filters, IEEE Signal Process.

Lett, vol.8, no. 1,pp.1 3,Jan 2001.

12 V.Crnojevi, V.enk, and Trpovski,

Advanced impulse detection based on pixel- wise MAD , IEEE Signal

Process,Lett,vol.11, no.7,pp.589 592 ,Jul. 2004.

13 R.H. Chan, C. Hu, and M. Nikolova,

An iterative procedure for removing Randon- valued impulse noise ,IEEE Signal Process,Lett, vol. 11, no. 12,pp.921 924, Dec. 2004.

14 H.Yu, L .Zhao, and H.Wang, An

efficient procedure for removing random valued impulse noise in images, IEEE Signal Process,Lett, vol. 15,

no. 1,pp.922 925, Dec. 22,2008.

15 F. Estrada,D.Fleet and A.Jepson,

Stochastic image denoising , in BMVC,2009,PP.1 11.

16 J.F.Cai,R.H.Chan ,and M.Nikolova,

Fast two-phase image deblurring under impulse noise,J.Math.Imag.Vis., vol ,36,

no.1 ,pp.46 53, 2010.

17 P.Trahanias and A.N.

Venetsanopoulos, Vector directional filters a new class of multichannel image

process filters, IEEE Trans. Image process, vol.2, no. 4,pp.528 534,1993.

18 Yiqiu Dong, Raymond H. Chan, and

Shufang Xu, A detection statistic for Random -Valued Impulse Noise.

19 Y.Dong and S.Xu, A New Directional

Weighted Median Filter for removal of random -valued impulse noise , IEEE

Signal Process Letter, vol.14,no.3,pp.193-196, Mar.2007.

20

Roman Garnett, Timothy Huegerich, Charles Chui, Fellow ,IEEE ,and W.He,A

Universal Noise Removal Algorithm with an Impulse Detector.