Applications Of Alternating Direction Method Of Multipliers In Image Denoising

The process of image denoising can be understood as the degradation process of image.This process is to recover a clear image from an image contaminated by noise as much as possible.There are many methods for image denoising,among them,the total variational denoising algorithm based on the theoretical basis of variational method is favored by related scholars because it can effectively describe the mathematical properties of images.The main attention in this thesis is focused on the applications of alternating direction method of multipliers in image denoising of total variation.This thesis presents two regularized denoising algorithms: a regularized denoising algorithm based on compressed sensing total variation and a regularized denoising algorithm based on adaptive directional total variation.The regularized denoising algorithm based on compressed sensing total variation first uses a total variation spectral framework method to decompose the noisy into high-frequency and low-frequency.This algorithm uses the prior knowledge of the frequency to regularization the image,and be improved the denoising ability on high-frequency parts of the image.Then,the alternating direction method of multipliers is used to solve this algorithm,and adapt the parameters of the fitting term.Finally,numerical experiments prove that the regularized denoising algorithm based on compressed sensing total variation can effectively denoise the additive noise of grayscale images.Compared with the comparative experimental algorithm,the peak signal-to-noise ratio is improved by more than 1d B,and the structural similarity is also better than comparative experiments.Experimental results show that this algorithm is suitable for removing images with large noise.The regularized denoising algorithm based on adaptive directional total variation uses prior knowledge of image direction to regularize the algorithm.First of all,the noisy image need be decomposed by a total variation spectral framework method.Then,the difference operator of the regular term in the algorithm is regularized by the image angle information.Among them,for the boundary artifact problem of the telescopic operator,the boundary weight of the parameters of the telescopic operator is improved to protect the edge processing information of the image.Subsequently,the algorithm is solved by the alternating direction method of multipliers.Finally,numericalexperiments prove that the regularization direction adaptive total variation denoising algorithm can effectively denoise the additive noise of grayscale images.Compared with the comparative experimental algorithm,the algorithm has better peak signal-to-noise ratio and structural compatibility.Experimental results show that this algorithm is suitable for denoising images with less noisy images and directional structures.