 Open Access
 Total Downloads : 904
 Authors : Jasmeen Kaur, Jagroop Singh
 Paper ID : IJERTV1IS7179
 Volume & Issue : Volume 01, Issue 07 (September 2012)
 Published (First Online): 25092012
 ISSN (Online) : 22780181
 Publisher Name : IJERT
 License: This work is licensed under a Creative Commons Attribution 4.0 International License
Image DeNoising Technique Based On Fuzzy Histogram Adaptive Filter
Jasmeen Kaur Jagroop Singh
Lecturer RBIENT Associate Professor Hoshiarpur DAVIET, Jallandhar
ABSTRACT
Noisy images require denoising techniques to filter out noise and produce noise free clear image. Noise in the image compromises the details of image. Many image processing techniques require removal of noise as a pre processing step. Huge amount of literature is present that deals with techniques to remove the noise from digital images. There are some traditional filters, some filters derived from traditional filters and some filters are new innovations. The work proposed here is related to the development of a novel fuzzy histogram adaptive filter to reduce noise from digital images. Experimental results show that the proposed technique is superior to other wellknown classical techniques available for the same purpose.
Keywords
Image denoising, spatial filters, fuzzy filters, mean filter, median filter, HAF

Introduction
Noise reduction from digital images is a very important and a wellresearched concept. Most of the image processing tasks require a clear or nonnoisy image to produce good results. Thus, noise reduction also acts as a preprocessing step for other image processing techniques.
Digital image processing deals with retrieving and manipulating discrete image data as per the specific application needs. This field is becoming more and more popular these days due to its vast applications ranging from medical image processing to satellite imagery and many others. The highlight of digital image processing is its versatility. Most of the applications of image processing seek an image which is free of noise due to various factors. Providing an error free image to these systems is a challenge. Many techniques have been proposed to free an image of different types of noise. Plenty of literature is available which deals with this problem.
Noise comes up during the process of acquiring the image from the source like sensor; digital camera etc.
Disturbances may occur due to ambient weather conditions or some other interference while capturing the image. The net effect is a corrupted image that needs to be preprocessed to reduce or eliminate the noise. Although noise gives an image a generally undesirable appearance, the most significant factor is that noise can cover and reduce the visibility of certain features within the image. The loss of visibility is especially significant for lowcontrast objects. In addition, sometimes images are not of good quality, due to both hardware and
software inadequacies, thus, they have to be enhanced and improved before other analysis can be performed on them. Figure 1 shows a sample noisy lena image along with its denoised version.
Figure 1. Sample Lena image with 5% random noise and denoised picture.
Before processing the image we have to remove the unwanted data from it, it means we have to remove the noise from the image. There are different types of noises each have different statistical properties [1].Various techniques are used to remove noise from image to enhance them. Some of the techniques are directly apply on the image data and some uses the transformations like wavelet or frequency. This paper presents fuzzy logic based technique to remove noise from digital images.

Literature Survey
Traditional filters include filters which are traditionally used to remove noise from images. We further divide
traditional filters into two classes: – spatial domain and transform domain.
In spatial domain various filters like mean filter[1], median filter, all work directly on the input image. It means it directly work on the pixels of original image. Out of various spatial based filters, wiener filter[2] gives better performance in case of Gaussian, Poisson and speckle noise. For impulse noise, median filter[4] outperforms all other filters. Various enhanced median filters like weighted median filter[5] are also used for this purpose.
It is needed when it is necessary to analyse the signal. Here, we transform the given signal to another domain and do the denoising procedure there and afterwards inverse of transformation is done in order to get final output. There are several transforms available like the Fourier transform, Hilbert transform, wavelet transform, etc. The Fourier transform is probably the most popular transform. Among different Fourier transforms[1] FFT is considered as best. However the Fourier transform is not give high performance in case of image denoising. Wavelet transform is better for this purpose[3]. Wavelet transform further provide different methods for removing noise from image which includes thresholding, non orthogonal wavelet transform and coefficient model.
Fuzzy based filters are those filters which includes concept of fuzzy logic in their filtering procedure. Fuzzy based filters can also be further classified into two categories: fuzzy classical and fully fuzzy.
Fuzzy classical filters include the filters which extend the traditional filters using fuzzy logic. There are plenty of fuzzy traditional filters on which many researchers have worked. We here mention only some of them. Popular fuzzy classical filters are Fuzzy median filter (FMF)[8], Fuzzy Impulse noise Detection and Reduction Method (FIDRM)[6], Fuzzy random impulse noise reduction method (FRINR)[7], Fuzzy weighted mean (FWM)[9], Adaptive weighted fuzzy mean (AWFM)[10]. In fuzzy median filter[8] and fuzzy weighted mean filter[9], fuzzy logic is added to enhance the traditional median and mean filters. Fuzzy impulse noise detection and reduction method[6] and Fuzzy random impulse noise reduction method[7] is two step methods. In first step noisy pixels are detected from the input image and after detection procedure, noise is removed from the detected pixels, this forms second step. The fuzzy logic is used in detection step by forming the fuzzy rules to decide whether the pixel is corrupted with noise or not. In filtering procedure, traditional filters like mean filter, median filter, weighted mean filter etc. are extended using fuzzy logic.
Fully fuzzy filters are those Denoising filters which are purely based on fuzzy logic and have no connection with traditional methods. Various methods that come under
this category are Fuzzy inference ruled by elseaction filter (FIRE)[11], dual step fuzzy inference ruled by else action filter (DSFIRE)[12], Piecewise Linear Fuzzy Inference Ruled by Elseaction Filter (PWLFIRE)[13], Gaussian noise reduction filter (GOA)[14], histogram adaptive filter (HAF)[15]. These are not all developed filters. We only mention some of the popular filters only. FIRE[11] filters are a family of nonlinear filters which adopt fuzzy rules toprocess image data.Dual Step FIRE filter[12], adopts fuzzy reasoning at two different levels in order to cancel noise pulses without damaging fine image structures. DSFIRE filter is able to largely outperform other methods in the literature. All fuzzy based filters use fuzzy rules based systems to add fuzzy logic in it.

Fuzzy Based Histogram Adaptive filter
The proposed histogram adaptive fuzzy (HAF) filter is particularly effective for removing highly impulsive noise while preserving edge sharpness. This is accomplished through a fuzzy smoothing filter constructed from a set of fuzzy IFTHEN rules, which alternate adaptively to minimize the output mean squared error as input histogram statistics change. An algorithm is developed to utilize (corrupted) input histogram to determine parameters for the nearoptimal fuzzy membership functions. Construction of the HAF filter involves three steps: (1) define fuzzy sets in the input space, (2) construct a set of IFTHEN rules by incorporating input histogram statistics to form the fuzzy membership functions, and (3) construct the filter based on the set of rules. Unlike many neurofuzzy or fuzzy neuro filters, where a random strategy is employed to choose initial membership functions for subsequent lengthy training, HAF can achieve nearoptimal performance without any training.

Proposed Technique
For the problem of interest here, we assume an input gray image sized 512Ã—512 with a pixel intensity from 0 to
255. With adaptive histogram approach i.e. Histogram Adaptive Filter (HAF), a fuzzy logic based technique was proposed to remove noise from digital images. In this technique we decide whether a pixel is noisy or not based on the intensity of the pixels surrounding the given pixel. For the said task we design a fuzzy inference system (FIS) that takes two inputs and produces one output as shown in Figure 2. The first input taken by the FIS is difference of the centre pixel and the pixel vertically/ horizontally preceding it. Second input the centre pixel and the pixel vertically/ horizontally succeeding it. We used Matlab for designing the proposed tool. Figure 3 shows membership function editor for the FIS.
Figure 2. FIS with input and output
Figure 3. Membership function editor
Figure 4. Rule editor
Figure 4 shows the rule editor for Matlab tool. The pixel is considered a noisy pixel if input 1 and input 2 i.e.; the differences of centre pixel with its neighbouring pixels is higher. Following are the rules considered for FIS:
If inp1 is very low and inp2 is very low then output is low
If inp1 is low and inp2 is low then output is low
If inp1 is med and inp2 is med then output is low
If inp1 is high and inp2 is high then output is high
If inp1 is very high and inp2 is very high then output is high
If inp1 is very low and inp2 is low then output is low If inp1 is very low and inp2 is med then output is low If inp1 is very low and inp2 is high then out is low
.
.
.
.
.
If inp1 is very high and inp2 is med then out is high
If inp1 is very high and inp2 is high then out is high
In case of the centre pixel which belongs to a one line thick line, the FIS may signal a noisy pixel. To avoid such situations, a three layered technique is used in which the output of first layer is fed to second layer and so on.

Results
We conducted extensive experiments on synthetic as well as natural images and computed PSNR value for each. The PSNR is then compared with mean filter results as well as median filter results. Following are the results of the same.
TABLE 1: COMPARATIVE RESULTS IN PSNR (DB) OF DIFFERENT FILTERS FOR LENNA IMAGE CORRUPTED WITH SALT AND PEPPER NOISE
NP %
Filtering Techniques
Mean Filter
Fuzzy Filter
Fuzzy based
HAF
1
35.1340
41.3238
42.4772
2
34.1683
40.8069
42.2556
3
33.4115
40.5509
41.9061
4
32.3040
39.7077
40.5690
10
30.2081
38.3102
39.8069
20
27.6814
35.3673
38.2126
30
26.1552
33.4299
36.1271
40
24.9798
31.5343
33.9347
50
24.0438
30.1847
33.1339
60
23.1996
28.5489
31.4157
70
22.4764
27.3113
30.2154
Figure 5. Restored results of enlarged Lena image corrupted with 40% of salt and pepper noise (a) True image, (b) Noisy image, (c) Mean Filtering, (d) Fuzzy filter (e) Fuzzy based histogram adaptive filter (HAF)

(b) (C)

(d) (e)
Figure 6. Performance comparison of different filtering methods on image degraded by salt and pepper noise

Conclusion
It is clear from the results section that the proposed technique outperforms the wellknown denoising filters called Mean filter and fuzzy filter. The proposed technique, as it is based on fuzzy logic, is well versed in handling the vagueness of information contained in noisy digital images.
References

Rafael Gonzalez Richard Woods, Digital Image Processing, Pearson Publications

A.K.Jain,Fundamentals of digital imageprocessing. PrenticeHall,1989

David L. Donoho and Iain M. Johnstone,Idealspatial adaption via wavelet shrinkage,
Biometrika,vol.81, pp 425455, September 1994.

N. C. Gallagher Jr and G. W. Wise, "A theoreticalanalysis of the properties of median filters", IEEE Trans.Acoust., Speech, Signal Processing, vol. ASSP29, pp.11361141, Dec. 1981.

R. Yang, L. Yin, M. Gabbouj, J. Astola, and Y.Neuvo, Optimal weighted median filter understructural constraints, IEEE Trans. Signal Processing,vol. 43, pp. 591604, Mar.1995.

Schulte S., Nachtegael M., De Witte V., Van der Weken D. &KerreE.E., A Fuzzy Impulse Noise Detection and Reduction Method in: IEEE TRANSACTIONS ON IMAGE PROCESSING, VOL. 15, NO. 5, MAY 2006, pp. 11531162.

Stefan Schulte, Valerie DeWitte, Mike Nachtegael, Dietrich Van derWeken, Etienne E.
Kerre, Fuzzy random impulse noise reduction method, in: Fuzzy Sets and Systems, ESLEVIER ,2006, pp. 270283.

Arakawa K., Median filter based on fuzzy rules and its application to image restoration, in: Fuzzy Sets and Systems, Vol. 77, 1996, pp. 313.

Arakawa K., Fuzzy rulebased image processing with optimization, in: Fuzzy Techniques
in Image Processing (Kerre E.E. &NachtegaelM.,editors), SpringerVerlag, 2000,
pp.222247.

Lee C.S., Kuo Y.H. & Yu P.T., Weighted fuzzy mean filters for image processing, in: Fuzzy Sets and Systems, Vol. 89, 1997, pp. 157180.

Russo F. &Ramponi G., A fuzzy filter for images corrupted by impulse noise, in: IEEE Signal proceedings letters, Vol.3, No. 6, 1996, pp. 168170.

Russo F. &Ramponi G., Removal of impulse noise using a FIRE filter, in: IEEE Proceedings, 1996, pp. 975978.

Russo F., FIRE operators for image processing, in: Fuzzy Sets andSystems, Vol. 103,1999, pp. 265275.

Van De Ville D., Nachtegael M., Van der Weken D., Kerre E.E., Philips W. &Lemahieu I., Noise reduction by fuzzy image filtering, in: IEEE Transactions on Fuzzy Systems, Vol11, Nr. 4, 2003, pp. 429436.

Wang J.H. & Chiu H.C., HAF: An adaptive fuzzy filter for restoringhighly corrupted images by histogram estimation, in: Proc. Natl. Sci.Counc. ROC(A), Vol. 23, No. 5, 1999. pp. 630643.

Naveen Kumar and Dr. Ramakrishna, An effeicient approach of removing high density salt and pepper noise using stationary wavelet transform, vol 12, march 2012.
