Efficient Removal of Noise from Image using Non-Linear Filtering

Abstract

An image may contain different types of information related to parameters such as colour and brightness, the irregular variation of which leads to noise in an image. Images often get corrupted by different kind of noise such as Impulse, Gaussian and Salt-and-Pepper Noise, during image acquisition or transmission caused by the digital sensor attempting to record/capture tiny amount of light. Since an image appears in multiple resolutions in different forms, an image can be segmented into different classifying features. As a result, signal and noise can be separated and elimination of noise becomes easier. In this article, a non-linear median filter is preferred, which is found to be useful in reducing not only Impulse Noise but also Gaussian and Speckle Noise. The corresponding analyses are performed on grayscale images, where the results obtained through non-linear filtering are superior to results obtained using Average, Gaussian and Weiner filters.