Image Denoising Using Multiresolution Analysis and Nonlinear Filtering

Abstract

In multiresolution analysis, image is analyzed at different resolutions of pixels. The features in an image appear at different resolutions in different forms. Hence separation of features in the image is possible. In denoising problems, signal and noise can be separated in the process and hence elimination of noise becomes easier. It is proposed in this paper that when a nonlinear median filter is used in multiresolution environment, once in full resolution and then with half resolution, denoising becomes more perfect. This technique is a non linear processing and is found to be useful in reducing not only impulse noise but also Gaussian and Speckle noise. Further, it is also proposed that use of a nonlinear adaptive median filter produces more pleasing image with better denoising. It is also shown that the proposed method is useful for color image denoising too.