Research And System Implementation Of Image Denoising Based On Total Variation Model

Image denoising is one of the indispensable steps before vision and subsequent analysis.At present,the research of efficient image denoising is still a key challenge.The variation method in digital image processing have always been favored by scholars at home and abroad because of its mathematical simplicity and the advantages of removing strong noise.In this paper,based on the study of the total variation model,new algorithms are proposed to solve the problems and difficulties existing in the existing models.The main contents of this thesis are as follows:(1)Aiming at the disadvantage that the total variation denoising model is easily affected by the gradient and the geometric features are easily lost,a weighted total variation image denoising algorithm based on Curvelet transform is proposed.Firstly,the noisy image is divided into n layers by a fast discrete Curvelet transform;then,combining the advantages of total variation model and Han's model,the weighted total variation model is obtained;then,according to the characteristics of Curvelet coefficients of each layer,the weighted total variation model is used for denoising;Finally,the images after denoising are reconstructed and the final denoised image are obtained.The experimental results show that the denoising effect of weighted total variation image based on Curvelet transform is better than that of traditional total variation denoising and non-local total variation denoising.(2)Aiming at the disadvantage that the total variation denoising model is easy to produce step effect and the weight parameters in different directions are identical,an adaptive weighted total variation image denoising algorithm is proposed to restore the image contaminated by noise.First,a new edge detector is constructed based on the similarity between the structure tensor and the non-local mean filter;then,use the new edge detector to construct an adaptive norm parameter;furthermore,horizontal and vertical weight parameters by adaptive means are selected adaptively;finally,the nonlocal mean filter is used to modify the fidelity term.The experimental results show that the adaptive weighted full variable image denoising algorithm proposed in this paper can keep the texture and edge of the image while removing the noise,and has a certain improvement in the subjective evaluation and objective evaluation.(3)Based on the above work,total variation image denoising system based on Matlab is designed and implemented.The system integrates the two algorithms proposed in this paper,as well as some classical denoising algorithms.