Two New Methods Of Gauss-salt And Pepper Mixed Noise Denoising

With the rapid development of digital products, there has been increasing requirementof the quality of images. However, images in reality often sufer from a variety of noisepollution. Usually, we only think about Gaussian white noise. It is a statistical noise andits probability density function is equal to that of the normal distribution. But in fact, theexistence of outliers (impulse noise) in the image data is not rare and can be caused by manyfactors.In this paper, we add salt-and-pepper, a kind of impulse noise, into the image with gaus-sian white noise, and we propose two novel algorithms to remove diferent degree of mixedGaussian-impulse noises. In our first algorithm, we minimize total variation with Huberpenalty function to get a preliminary denoised image, and based on this, we modify Block-Matching (BM3D) method to get the final denoised image. In the second algorithm, Wefirstly utilize sparse and low rank matrix decomposition to remove most of the impulse noiseand obtain our final denoised image by Block-Matching and collaborative wiener filtering.Our numerical results show that the two methods can remove the mixed noise efectively.Compared with other methods, they also can deal with strong Gaussian noise.