Research On Image Denoising Algorithms Of Two Types Of Non-Gaussian Noises

In image processing,noise caused by sensors or other interference factors is usually modeled as Gaussian white noise.However,in practical applications,the noise may be non-Gaussian or even impulsive noise with heavy-tailed distributions.This paper mainly discusses the denoising problems of two types of non-Gaussian noise:Random-Valued impulse noise and Symmetric alpha-stable distribution noise,and derives two image denoising algorithms based on maximum a posteriori（MAP）estimation and total variation（TV）model.The innovation of this paper lies in the design of the fidelity terms of the two algorithms,as well as the proposed approximation method for probability density function of non-special cases of Symmetric alpha-stable distribution.For the removal of Random-Valued impulse noise,the maximum concave penalty（MCP）function is chosen as the fidelity term of the model,considering the characteris-tics of random-valued impulse noise and the property of non-convex functions.Com-pared with norms like L₁and L₂,the MCP function can better approximate the L₀norm and can be more compatible with the solving problem by controlling its parame-ters.For the removal of Symmetric alpha-stable distribution noise,this paper approx-imates the probability density function of Symmetric alpha-stable distribution,derives the algorithm under the framework of maximum a posteriori estimation,and proposes an estimation method for noise-related parameters of a given Symmetric alpha-stable distribution noisy image.The main algorithm for solving the model in this paper is the Alternating Direction Multiplier Method（ADMM）,and the convergence of two algo-rithms are analyzed under certain conditions.The numerical experiments show that the proposed model outperforms some existing traditional restoration methods in terms of both quantitative evaluation indicators and visual quality.