International Journal of Computer Applications (0975 – 8887) Volume 109 – No. 10, January 2015 18 Image Transmission over Multipath Fading Channel and Image Denoising using Directional Weighted Median Filter Veeramma Yatnalli JSS Academy of Technical Education, Bengaluru Karnataka India K L Sudha Dayananda Sagar College of Engineering, Bengaluru Karnataka India ABSTRACT Impulse noise is often introduced into images during transmission contaminating the images due to channel errors. Based on the noise values, impulse noise can be classified as Fixed Valued Impulsive Noise or Salt and Pepper Noise(SPN) and Random Valued Impulsive Noise(RVIN). In this paper, an inpainting algorithm is presented based on Directional Weighted Median(DWM) Filter to denoise both the noises caused due to image transmission over multipath fading channel. The algorithm diffuses median value of pixels from the exterior area into the inner area and thus preserves the edges and fine details. The random valued impulsive noise and salt and pepper noise due to wireless channel modeled are simulated using MATLAB channel objects. The detection algorithm combined with the image correction based on DWMF shows better performance in terms of Peak Signal-to- Noise Ratio(PSNR) and Mean Absolute Error(MAE). General Terms Image Reconstruction, Channel Fading Keywords Inpainting, Random Valued Impulsive Noise(RVIN), Salt and Pepper Noise, Directional Weighted Median(DWM), PSNR, MAE. 1. INTRODUCTION When image and video are transmitted over noisy channels, the data is either missing or incorrect due to channel transmission errors. As a result, impulse noise can appear because of a random bit error on a communication channel. The resulting two noises that corrupt the source images based on the noise value are:  Salt and Pepper impulse Noise, which means a noisy pixel has a high value due to positive impulse noise and looks like white dot or snow in the image, or has a low value due to a negative impulse noise and looks like black dot or pepper in the image.  Random Valued Impulse Noise can take any value that do not occur as extreme outliers in comparison with the surrounding pixels. Nonlinear filters, which are based on statistical ordering of pixel values in a predefined fixed size sliding window, are effective in suppressing impulse noise in images. However, when they are applied to an image having uniformly distributed image values, undesired processing of noise-free pixels results in edge and texture blurring. The de-noising techniques based on median value works fine for restoring the images corrupted by Random Valued Impulse Noise with low noise level but exhibits poor performance with highly corrupted images. To achieve an optimal balance between signal-detail preservation and impulse noise attenuation, several works have been done to investigate impulse detector algorithms prior to noise removal techniques. These impulse detector techniques address the challenge of detecting noisy samples and selectively apply the result of noise removal techniques only to affected image regions. Many restoration algorithms have been proposed in the past to address the problem of filling in missing/corrupted data. Image Inpainting is one such technique which restores the damaged pixels in an undetectable form. Median based Inpainting technique diffuses the median of pixels which are exterior to inpainting area into the area to be inpainted. This technique is simple which achieves better results than other Inpainting techniques. Partial Differential Equation (PDE) based algorithms [1] are designed to connect edges and discontinuities and iteratively propagates information from outside of the area along isophotes. In convolution and filter based methods, inpainting is done by convolving the damaged image with a proper kernel. In this paper, Oliveira [2] proposed a fast digital image Inpainting method, which depends on convolution operation. This approach convolves regions to be inpainted with a diffusion mask repeatedly. However, convolution and filter based algorithms have good results only for images having no high contrast edges or high frequency components. When an image has high contrast edges or high frequency components, these algorithms produces some blurring in edge regions. In addition, the above techniques fail to recover the large regions containing the edges and are applied for the entire image. In the following paragraph, papers are referred for the purpose of studying median based techniques to restore the pixels damaged due to image transmission over fading channel. A large number of algorithms have been proposed to remove impulse noise. In paper [3], starting from basics of median filter, the theory of Weight Median filtering and up to recently developed theory of optimal weighted median filtering for speech processing, adaptive weighted median and optimal weighted median filters for image and image sequence restoration, weighted medians as robust predictors in DEP coding and Quincunx coding is available. In paper [4], the directional median filter performs one-dimensional median filtering in the direction of the straight line component detected around the scratch. This method works effectively, even if a similar undamaged image area is not available. Among the works [5], [6], the median based filters are widely used because of their effective noise suppression capability and high computational efficiency. However, the