Image denoising using local adaptive layered Wiener filter in the gradient domain

This paper presents a new image denoising algorithm. Our method is inspired by locally adaptive window-based denoising using maximum likelihood (LAWML). In the research, we find, as with wavelet coefficients, the gradient image coefficients can also be modeled as zero-mean Gaussian random variables with high local correlation. So, we implement the local adaptive Wiener filter in the gradient domain. But unlike LAWML, the layered denoising is adopted in our method. At the same time, the relation between wavelet-based and diffusion-based denoising method is disclosed further. The tests demonstrate the proposed method gets the desired results both subjectively and objectively compared to the related gradient domain algorithms and wavelet-based image denoising methods. At the same time, the tests also show the proposed method outperforms some other diffusion filters and wavelet-based methods and Non-Local means (NL-means) filter in most cases.