Research On Mixed Noise Removal Based On Laplace Distribution

In modern digital image processing and computer vision applications,image denoising is always an important problem.Noise is inevitably introduced in the process of image acquisition,transmission and storage,among which,Gaussian noise and saltand-pepper noise are the two most common noise types,which will destroy the details and edge information of the image and make the image quality decline.Traditional Gauss-salt mixed noise removal algorithms use regularization terms to constrain the model,while their computational complexity is high.The methods based on neural network have strong learning ability but lack interpretability.In this paper,we propose a mixed noise removal method based on Laplace distribution and plug-and-play.In the first place,the noise is modeled with Laplace distribution and the problem is established under the framework of maximum a posteriori.Then alternating direction multiplier method is used to decompose the problem,after that we can insert an off-line pretrained Gaussian denoiser to solve one of the sub-problems,which is also known as plug-and-play.The validity of the algorithm is proved on the standard test set.Under the proposed framework,other Gaussian denoisers can also be inserted.Furthermore,we extend the Laplace distribution to generalized Laplace distribution,and use generalized soft threshold operators and grid search to solve new subproblems.For different images with different noise levels,the generalized Laplace distribution can describe the noise more accurately,and further optimize the denoising effect.