The Research Of Image Denoising Algorithm Based On Nonlocal Means And Total Variation Model

Image denoising technology is an important part of computer vision and image processing.In real life,the image is easily polluted by noise.A good image denoising method can improve the image quality and increase the accuracy of subsequent image processing and analysis.At present,for the removal of noise in images,although a lot of noise reduction models have been proposed,but these denoising models often have the problem of the uncertain parameters and the type of denoising is single.The more in-depth analysis of these denoising algorithms can improve the application value of the algorithm,and has very practical significance.The nonlocal means(NLM)algorithm and the total variation(TV)model are two very classic image denoising methods.This paper will analyze and study image denoising algorithm based on these two algorithms respectively.This paper first analyzes the factors affecting the denoising performance of the two methods;then for the problem of the weight function's smooth parameter of the NLM algorithm is not clear,according to the gaussian noise distribution rule,proposed a new calculation model of the weight,and gives the corresponding calculation method for weight parameters adaptively;finally for the problem of TV model can not effectively remove impulse noise,the value of fidelity parameters of the model are modified,according to the different type of the noise contained in the pixels,assigning different fidelity parameter values for each pixel,so it has the ability to remove impulse noise and Gaussian.Through the comparison of experimental results and data analysis,to prove the rationality of the new weights model,the validity of the weight parameters calculation strategy and the correction method for TV model,to achieve the desired purpose.