Research On Adaptive NLM Image Denoising Algorithm Based On Regularization Model

Image denoising is still one of the most basic and important tasks in image follow-up operations,such as image restoration,visual tracking,and image segmentation.Nonlocal Means(NLM)image denoising reduces noise based on redundant information of images in the pixel spatial domain,and its denoising effect is better.Traditional NLM image denoising still has some problems,such as using the same size search window to denoise,ignoring the difference between the pixel in the smooth region and the texture region,and the improvement of the NLM weight kernel function and image structure similarity has been relatively Mature,the space for improving the denoising effect is limited.Aiming at the above problems existing in the original NLM,two improved schemes for NLM image denoising are proposed.The details are as follows:(1)The development process of image denoising algorithm and the research status at home and abroad are briefly introduced.The core algorithm of image denoising is represented by mathematical model.Several classical image denoising algorithms are introduced in detail.The algorithm for denoising NLM image is explained.Principle;classify noise and describe it with mathematical model;classify noise and describe it with mathematical model;divide image evaluation criteria into subjective and objective;provide theoretical basis for subsequent research.(2)The improvement of the NLM weight kernel function and the similarity of image structure is relatively mature,and the space for improving the denoising effect is limited.Aiming at this problem,based on the traditional NLM algorithm with good denoising effect,an NLM image denoising algorithm based on regularization model is proposed.The regularized denoising model is derived by Bayesian Maximum A Posteriori Probability Estimation(MAP),and the regularization model is optimized by solving the block by the Alternating Direction Method of Multipilers(ADMM).By adding different levels of noise to the traditional NLM algorithm and the proposed algorithm,the experimental results show that compared with the traditional NLM algorithm,the proposed algorithm not only retains more image features in visual aspect,but also denoises.The performance has been further improved.(3)The traditional NLM algorithm uses the same size of Search Area(ROS)to denoise the image,and its performance decreases as the noise variance increases.If the pixel is in a smooth region,considering the contribution of more similar pixels in the averaging process,the size of the search window should be large.Similarly,if the pixel is in a non-smoothed area,the search window area should be small.Therefore,an adaptive NLM image denoising algorithm based on image search region is proposed.The method uses the robust median estimation method in the wavelet domain to estimate the image noise.Corresponding to three different search regions through noise estimation,the search window is divided into three levels: large,medium and small.The experimental results show that the proposed algorithm is superior to the traditional NLM image denoising method in objective evaluation and image visual effects.