A PROPOSED METHOD OF SALT AND PEPPER NOISE REMOVAL IN IMAGE PROCESSING

Abstract

Salt and pepper denoising is an important pre-processing step in image processing to help improve image quality. Traditional denoising methods consider pixels with zero or maximum values as actual salt-and-pepper noise pixels. However, in many cases, these pixels are not noise pixels but rather texture pixels that lead to limited denoising effectiveness, especially when the noise ratio is high. This study overcomes the above limitation. At the noise pixel detection step, the algorithm uses an adaptively sized window to help distinguish the noise candidate points as real noise pixels or texture pixels based on the observation that the noise pixels are usually isolated pixels, whose value changes suddenly compared to the surrounding values. Then the noise pixel will be replaced by a new value based on the value of the neighboring pixels in the window. This method is only applied to grayscale images. The experimental results give the mean peak signal-to-noise ratio and structural similarity at all noise levels of 28.1631 and 0.8496, respectively. The proposed method for removing salt-and-pepper noise is very effective based on experimental results. The result is good quality images at all noise levels.