Image Denoising Algorithm Based On Nonlocal Variational

As a very important information carrier,image has been used in various fields widely,such as computer vision,intelligent transportation,industrial manufacturing,aerospace and so on?However,image will inevitably be disturbed by various kinds of noise in the process of acquisition,storage and transmission,that will result in the distortion of the image signal.But in today's life,high-quality images are required in many fields,image which is disturbed by noise is difficult for image application.So it is necessary to denoise the noisy image.As an important mathematical theory,the variational method has extensive application and research in image anti shake and image defuzzification.In the field of image denoising,variational method combines the redundancy and non locality of image,and establishes the nonlocal variational model of noisy image to solve the image denoising problem.Nonlocal variational model has a good theoretical basis and practical denoising effect,so it has great research value.Based on the nonlocal variational method,this paper studies its application in image denoising,analyzes the advantages and disadvantages of it,and put forward optimization application and improvement according to the characteristics and principles of the nonlocal variational method.The main work and research results of this paper are as follows:1.Summarize the domestic and foreign research status of image denoising and the mainstream image denoising algorithm.NLM?NLTV?BM3D image denoising algorithms are analyzed and studied.The shortcomings of these algorithms and the problems to be solved are analyzed.2.An image denoising algorithm based on ICA and NLTV algorithm is proposed.Experiments show that the proposed algorithm(NLTV-ICA)gets an better performance than NLM?NLTV in visual effect and objective evaluation3.An new adaptive NLTV(NA-NLTV)model is proposed.The NA-NLTV model optimizes the balance between fidelity term and regularization term by creating an adaptive balance function which can identify the texture information of the image according to the redundancy and similarity of the image.The experiment shows that the NA-NLTV model has better denoising effect.4.In order to solve the problem that the denoising algorithm proposed in this paper has larger time complexity which results in slower running speed,this paper uses GPU parallel computing to accelerate the speed of NLTV-ICA and NA-NLTV,which greatly improves the speed of our algorithm.