A Novel Color Image Denoising Technique Using Window Based Soft Fuzzy Filter Hemant Kumar, Dharmendra Kumar Roy Abstract - The image corrupted by different kinds of noises is a frequently encountered problem in image acquisition and transmission. The noise comes from noisy channel transmission errors. The impulse noise (or salt and pepper noise) is caused by sharp, unexpected disturbances in the image signal; its appearance is randomly scattered white or black (or both) pixels over the image. Gaussian noise is an idealized form of white noise, which is caused by some random fluctuations in the signal. Speckle noise (or more simply just speckle) can be modelled by random values multiplied by pixel values; hence it is also called multiplicative noise. This work presents a novel technique for edge preserved color image denoising using window based soft fuzzy filter based on asymmetrical triangular membership function. However lots of techniques like median, mean and average filters are available for gray image denoising, but most of the time it is found that all these filters are capable to provide good noise removal for some specific type of noise, but cant able to preserve the edges of the images ie the output images were greatly suffers from the blurring effect. So to address this problem the proposed technique not only concentrates on efficient noise removal as well as preservation of image edges. To handle this problem fuzzy logic based soft technique is proposed, because of imprecise and vague situations handling capability of fuzzy based techniques. To illustrate the proposed method, experiments have been performed on color test image like Lena and results are compared with other popular image denoising methods. For the comparative analysis of the proposed work a comparison between conventional filters and proposed filter has been also presented in the thesis on the basis of three important parameters Mean square error (MSE), Peak signal to noise ratio (PSNR) and Edge Preservation index (EPI). The obtained results show that the proposed method has very good performance with desirable improvement in the PSNR and MSE of the image. Index Terms Color Image denoising, edge preservation, fuzzy filters, membership function, triangular membership function, PSNR, MSE, EPI (Edge Preserving Index). I. INTRODUCTION The fundamental problem of image and signal processing is to effectively reduce noise from a digital image while keeping its features intact (e.g., edges). Three main types of noise exist: impulse noise, multiplicative noise and additive noise. Impulse noise is usually characterized by some portion of image pixels that are corrupted, leaving the remaining pixels unaffected. Examples of impulse noise are fixed valued impulse noise and randomly valued impulse noise. The additive noise comes when a value from a certain distribution is added to each one image pixel, for example, a Gaussian distribution. Multiplicative noise is usually more difficult to remove from images than additive noise because the intensity of the noise varies with the signal intensity (e.g., speckle noise). Fuzzy set theory and fuzzy logic [1] offer us powerful tools to represent and process human knowledge represented as fuzzy rules. Fuzzy image processing [2] has three main stages: 1) Image Fuzzyfication, 2) Modification of membership values, and 3) Image Defuzzyfication. The Fuzzyfication and Defuzzyfication steps are due to the fact that we do not yet possess fuzzy hardware. Therefore, the coding of image data (Fuzzyfication) and decoding of the results (Defuzzyfication) are steps that make it possible to process images with fuzzy techniques. The key power of fuzzy image processing lies in the second step. After the image data is transformed from input plane to the membership plane (fuzzyfication), suitable fuzzy techniques modify the membership values. It can be a fuzzy clustering, a fuzzy integration approach, a fuzzy rule-based approach,, etc. This paper presents a novel technique for edge preserved color image denoising using window based soft fuzzy filter based on asymmetrical triangular membership function. In this work, the input image is first divided into three Red, Green and blue constituent single channels and then a fuzzy membership-type of weighted functions is applied [2] to each single channel image pixel-values within a moving window, and define a fuzzy filter based on asymmetrical membership function. This fuzzy filter attempt to incorporate the feature of a moving average filter for filtering noise and also preserves the edges of the image. Obtained results shows that this fuzzy filter have great success in filtering images with random noise, impulse noise, Gaussian noise and speckle noise. II. DEFINITION OF FUZZY LOGIC BASED SOFT FILTERS Let x (i, j) be the input of a 2-dimensional fuzzy filter, the output of the fuzzy filter is defined as: ISSN: 2278 1323 All Rights Reserved 2014 IJARCET 2789
y(i, j) = (r,s) A F x i+r,j+s.x(i+r,j+s) (r,s) A F[x i+r,j+s ] (1) Where F[x(i, j)] is the window function and A is the area of the window. With different window functions, we now define a novel fuzzy filter, which we shall call the fuzzy logic based soft filter with asymmetrical triangular membership function with moving average center (FLBSFWATMF). Start Read Multichannel (Color) Input Image A. FUZZY LOGIC BASED SOFT FILTER WITH ASYMMETRICAL MEMBERSHIP FUNCTION Add Noise to the color input image The Fuzzy Logic Based Soft Filter with an asymmetrical triangular function and the moving average value within a window as the center value is defined as: Extract Red s of noisy input image Extract Green s of noisy input image Extract Blue s of noisy input image (2) The degree of asymmetry depends of the difference between xmav(i, j) - xmin{i, j) and xmax (i, j)-xmav(i,j). xmin (i, j) and xmav(i, j) represent, respectively, the maximum value, the minimum value, the moving average value of x(i+r, j+s) within the window A at discrete indexes (i, j). Obtain values of Xmin, Xmav and Xmax for Asymmetrical Membership function for Red of input color image. Obtain values of Xmin, Xmav and Xmax for Asymmetrical Membership function for Green of input color image. Obtain values of Xmin, Xmav and Xmax for Asymmetrical Membership function for Blue of input color image. III. METHODOLOGY The idea of this proposed work is to average a pixel using other pixel values from its neighborhood, but simultaneously preserve edges of the image which should not be destroyed by the filter. The complete methodology of the proposed is shown in figure (3.1) with the help of flow chart representation. In this project we take input multichannel color image after then we add noise like Gaussian noise or Speckle noise. We extract three color images Red, Green, Blue and after then obtain the value of Xmin, Xmax and Xmav for Asymmetrical Membership function for Red, Green and Blue images. Then we initialize the moving window for fuzzy filter development for Red, Green and Blue. Then we apply Fuzzy filter in row and column wise on noisy Red, Green and Blue Component. Finally combine the three Red, Green, Blue s of filtered image to generate combined filtered color image as output and Display Filtered color image and Calculate MSE, PSNR, EPI. Initialize the moving window for fuzzy filter development for Red Apply fuzzy filter defined by eq. 1 in row and column wise on noisy Red Initialize the moving window for fuzzy filter development for Green Apply fuzzy filter defined by eq. 1 in row and column wise on noisy Green Initialize the moving window for fuzzy filter development for Blue Apply fuzzy filter defined by eq. 1 in row and column wise on noisy Blue Combine the three Red, Green & Blue Components of filtered image to generate combined filtered color image as output Display filtered color image and Calculate MSE, PSNR and EPI Stop Figure (3.1) Flow chart representation of proposed work. ISSN: 2278 1323 All Rights Reserved 2014 IJARCET 2790
IV. RESULTS This section presents complete visual and comparative analysis of salt and pepper noise denoising of noisy images using average filter, median filter and developed fuzzy based filter. Figure (4.1) shows first noisy input image i.e. lena.jpg, which is corrupted by 50% salt and pepper noise. To present a comparative performance evaluation of developed algorithm, obtained results will be compared with the median filter and average filter, which are the most efficient filter for salt and pepper noise removal. The results obtained after the denoising process using developed algorithm and conventional filters are shown from figure (4.2) to figure (4.4). For example Figure (4.2) shows the resultant image after denoising using average filter, Figure (4.3) shows the resultant image after denoising using median filter, and Figure (4.4) shows the resultant image after denoising using developed fuzzy filter. Table (1) shows the parameter values obtained after denoising using the three filters. Figure (4.3) Figure (4.4) TABLE I. Comparison of Various parameters wrt change in median, average and fuzzy filters for input image 1 Figure (4.1) S. No. Parameters Average Median Developed Fuzzy Filter 1 MSE 99.10 32.46 21.20 2 PSNR 29.07 33.02 34.94 3 EPI 0.17 0.25 0.73 From table 1, it is evident that the developed fuzzy filter provides least MSE and highest PSNR as compare to conventional filters. In addition to this it is also clear from the table, that the developed fuzzy filter provides highest amount of edge preservation during first image denoising. Figure (4.2) Similarly figure (4.5) shows second noisy input image ie. pears.png, which is also corrupted by 50% salt & pepper noise. The results obtained after the denoising process using developed algorithm and conventional filters are shown from figure (4.6) to figure (4.8). For example Figure (4.6) shows the resultant image after denoising using average filter, Figure (4.7) shows the resultant image after denoising using median filter, and Figure (4.8) shows the resultant image after denoising using developed fuzzy filter. Table (2) shows the parameter values obtained after denoising of second input color image using three filters. ISSN: 2278 1323 All Rights Reserved 2014 IJARCET 2791
International Journal of Advanced Research in Computer Engineering & Technology (IJARCET) Figure (4.7) Parameters Average Median Developed Fuzzy Filter 1 MSE 103.51 16.46 10.86 2 PSNR 29.30 35.97 37.81 3 EPI 0.22 0.25 0.69 S. No. Figure (4.8) TABLE II. Comparison of Various parameters wrt change in median, average and fuzzy filters for input image 2 Figure (4.5) From table 2 it is evident that the developed fuzzy filter provides least MSE and highest PSNR as compare to conventional filters. In addition to this it is also clear from the table that the developed fuzzy filter provides highest amount of edge preservation during first image denoising. Similarly figure (4.9) shows third noisy input image ie. Football.jpg, which is also corrupted by 50% salt & pepper noise. The results obtained after the denoising process using developed algorithm and conventional filters are shown from figure (4.10) to figure (4.12). For example Figure (4.10) shows the resultant image after denoising using average filter, Figure (4.11) shows the resultant image after denoising using median filter, and Figure (4.12) shows the resultant image after denoising using developed fuzzy filter. Table (3) shows the parameter values obtained after third image denoising using three filters. Figure (4.6) Figure (4.9) ISSN: 2278 1323 All Rights Reserved 2014 IJARCET 2792
conventional filters. In addition to this it is also clear from the table that the developed fuzzy filter provides highest amount of edge preservation during third image denoising. Figure (4.10) Figure (4.11) V. CONCLUSIONS In this modern era during image acquisition, transmission and reception, the images are highly influenced by different source of noises. Hence for proper interpretation of image information the images must de-noised in all the three stages. In this work a robust and efficient gray image denoising technique has been successfully developed using window based soft fuzzy filter with asymmetrical triangular membership function. Although good denoising techniques are already available for image denoising like median filter and average filter, while most of time it has been found that all these filters provide good results but not able to preserve image edges during denoising. Ie the resultant images from conventional techniques are highly blurred. Since edges are very important characteristics it must be preserved during denoising process. Section ~ 4 shows the results obtained after denoising of three images using developed technique, median filter and average filter. From the tables it is clearly evident that for all three images MSE obtained for fuzzy filter is less as compare to conventional filters for the two different types of noises, on the other side the PSNR value is higher for fuzzy filter as compare to conventional filter. Figure (4.12) TABLE III. Comparison of Various parameters wrt change in median, average and fuzzy filters for input image 3 S. No. Parameters Average Median Developed Fuzzy Filter 1 MSE 25.82 35.27 25.18 2 PSNR 34.31 32.66 34.12 3 EPI 0.12 0.12 0.38 In addition to this the most important task of the developed algorithm is to preserve the image edges during denoising process. From the result tables it is clear that the edge preservation index (EPI) for fuzzy filter is 50% higher than for median filter and average filter. For analysis purpose only salt & pepper noise has been utilized, in future this analysis can be extended for other type of noises. REFERENCES [1] E. E. Kerre, Fuzzy Sets and Approximate Reasoning. Xian, China: Xian Jiaotong Univ. Press, 1998. [2] H. R. Tizhoosh, Fuzzy-Bildverarbeitung: Einfhrung in Theorie und Praxis. Heidelberg, Germany: Springer-Verlag, 1997. [3] F. Farbiz and M. B. Menhaj, A fuzzy logic control based approach for image filtering, in Fuzzy Tech. Image Process., E. E. Kerre and M. Nachtegael, Eds., 1st ed. Heidelberg, Germany: Physica Verlag, 2000, vol. 52, pp. 194 221. [4] D. Van De Ville, M. Nachtegael, D. Van der Weken, W. Philips, I. Lemahieu, and E. E. Kerre, A new fuzzy filter for Gaussian noise reduction, in Proc. SPIE Vis. Commun. Image Process., 2001, pp. 1 9. [5] D. Van De Ville, M. Nachtegael, D. Van der Weken, E. E. Kerre, and W. Philips, Noise reduction by fuzzy image filtering, IEEE Trans. Fuzzy Syst., vol. 11, no. 8, pp. 429 436, Aug. 2003. [6] M. Nachtegael, S. Schulte, D. Van der Weken, V. De Witte, and E. E. Kerre, Fuzzy filters for noise reduction: The case of Gaussian noise, in Proc. IEEE Int. Conf. Fuzzy Systems, 2005, pp. 201 206. [7] D. Donoho, Denoising by soft-thresholding, IEEE Trans. Inf. Theory, vol. 41, no. 3, pp. 613 627, May 1995. From table 3 it is evident that the developed fuzzy filter provides least MSE and highest PSNR as compare to ISSN: 2278 1323 All Rights Reserved 2014 IJARCET 2793
[8] C. C. Lee, Fuzzy logic in control systems: Fuzzy logic controller-parts 1 and 2, IEEE Trans. Syst., Man., Cybern., vol. 20, no. 2, pp. 404 435, Mar. Apr. 1990. [9] J. Fodor, A new look at fuzzy-connectives, Fuzzy Sets Syst., vol. 57, no. 2, pp. 141 148, July 1993. [10] R. Garnett, T. Huegerich, C. Chui, and W. He, A universal noise removal algorithm with an impulse detector, IEEE Trans. ImageProcess., vol. 14, no. 11, pp. 1747 1754, Nov. 2005. [11] Amaninder Kaur Brar, Vikas Wassan, Image Denoising Using Improved Neuro-Fuzzy Based Algorithm: A Review, Volume 4, Issue 4, April 2014, ISSN: 2277 128X, 2014, IJARCSSE. [12] Min-Chi Kao; Chia-Hung Lin; Li, T.-H.S., "Ant colony optimization based fuzzy image filter design for removal of impulse noises," Advanced Robotics and Intelligent Systems (ARIS), 2013 International Conference on, vol., no., pp.98,103, May 31 2013-June 2 2013. [13] A. K. Chandrakar, R. Dewangan Development of Efficient Color Image Compression Technique using Modified JPEG 2000 Standard ijafrc, Volume 1, Issue 5, May 2014. ISSN 2348 4853. AUTHOR PROFILE Hemant Kumar is a M.Tech. Scholar of RCET, Bhilai (C.G.), India. He did his M.C.A. From Chhattisgarh Swami Vivekananda Technical University Bhiali, Chhattisgarh Mr. Dharmendra Kumar Roy is working as Reader in Department of Computer Science & Engg., RCET, Bhilai (Chhattisgarh), India. He has published much research paper in international journals and presented several research papers in international conferences. ISSN: 2278 1323 All Rights Reserved 2014 IJARCET 2794