Weighted Hybrid Wavelet Wiener Filter for Gaussian Noise Removal

Abstract

A novel technique capable of removing Gaussian noise with less computational complexity has been
presented. This paper proposes a hybrid filter which employs Wavelet Transforms, a very powerful multiresolution tool,
employing modified Bayes thresholding, in conjunction with the Wiener Filter. The Wiener filter tries to build an optimal
estimate of the original image by enforcing a minimum mean-square error constraint between estimate and original
image. In the first step, Discrete Wavelet transform is applied to the given image, using modified Bayes thresholding for
better performance. This is followed by application of Wiener Filter to the output obtained in the previous stage. The
proposed algorithm is tested on a number of benchmark images and is found to produce better results in terms of the
qualitative and quantitative measures of the image for both low and high values of noise variance in comparison to many
existing techniques. The proposed technique removes Gaussian noise and the edges are better preserved with less
computational complexity and this aspect makes it easy to implement in hardware.