Enhanced Adaptive Weighted Mean Filter With Non-Local Means To Denoise Image Corrupted By Salt And Pepper Noise

There exist different kind of techniques to remove unwanted signals from images. The major denoising methods include filtering technique which is under the category of neighbourhood methods.The drawback of neighbourhood method is that tend to lose fine details of the image so that blurring may occur.Noise reduction and preservation of actual image can be done efficiently if we extend the consideration area such a way that not only comparing the grey level in a single point but also the geometrical configuration in a whole neighbourhood.This paper combines a new adaptive weighted mean filtering technique and the nonlocal means filtering algorithm to denoise image corrupted by salt and pepper noise. Keywords— De-noising, Non-local means, Salt and pepper noise INTRODUCTION Image processing is applicable for improving the image quality. Image noise is the random variation of brightness or colour information in images. It can be classified as Salt and pepper noise and random valued noise. The salt impulse noise have the brightest gray level and appear as white spot as well as pepper impulse noise having darkest gray level with blak spot. The SPN is also called fat-tail distributed, impulsive noise or spike noise. This paper proposes a technique of two level filtering .The corrupted input image is first subjected to a new adaptive weighted mean filter. For each pixel, we firstly determine the changing window size by continuously enlarging the size up to the point in time the extreme values of two nearby windows are equal respectively. Then the current pixel is regarded as noisy pixel if it is equal to the maximum or minimum values, otherwise, it is regarded as noise-free pixel. The filtered image is then de-noised using nonlocal means filtering technique[2]. The NLM filter was introduced by Buades in 2005[1]. This method of image denoising consist of weighted average of all pixel intensities where the family of weights depends on the similarity between the pixels and the neighbourhood of the pixel being processed[3]. PROPOSED SYSTEM The main objective of my work is enhance the performance of denoising. For that a two phase filtering technique is proposed here. The image can be corrupted by different levels of salt and pepper noise. A. Noise Detection-Enhanced Weighted Mean Filter For each pixel ,we firstly determine the adaptive window size by continuously enlarging the window sizes up to the point in time the extreme values of two successive windows are equal respectively. Then the center pixel is regarded as noise candidate if it is equal to the maximum or minimum values, otherwise it is regarded as noise free pixel. By this way the detection error can be largely decreased especially for very high level noise[4]. B. Noise Removal-Nonlocalmeans Filtering The approach of Non Local Means filtering is based on estimating each pixel intensity from the information provided from the entire image and hence it exploits the redundancy caused due to the presence of same kind of patterns and features in the image. In this method, the restored gray value of each pixel is obtained by the weighted average of the gray values of all pixels in the image. The weight assigned is proportional to the similarity between the local neighborhood of the pixel under consideration and the neighbourhood corresponding to other pixels in the image. Given a discrete noisy image v= v(i) for a pixel i the estimated value of NL[v](i) is computed as weighted average of all the pixels NL[v](i)= w(i,j) v(j) where the family of weights w(i, j) depend on the similarity between the pixels i and j. The similarity between two pixels i and j depends on the similarity of the intensity gray level vectors v(Ni) and v(Nj ), where Nk denotes a square neighborhood of fixed size and centered at a pixel k. The similarity is measured as a decreasing function of the weighted Euclidean distance. The pixels with a similar grey level neighborhood to v(Ni) have larger weights in the average. These weights are defined as, where Z(i) is the normalizing constant and the parameter h acts as a degree of filtering. It controls the decay of the exponential function and therefore the decay of the weights as a function of the Euclidean distances[1]. Fig. 1. proposed system Sukina K et al, / (IJCSIT) International Journal of Computer Science and Information Technologies, Vol. 6 (5) , 2015, 4751-4752