Research On Denoising Of Non-local Mean Image Based On Edge Detection

With the advancement of science and technology,images have become important information carriers in our daily lives.The widespread use of images has promoted the vigorous development of the image technology.However,in practical applications,images are subject to noise pollution due to a series of causes such as inclement weather or electronic devices,resulting in a drop in image quality.In severe cases,it may no longer be possible to continue use.Therefore,with the promotion of image processing,how to remove the noise in the image has also received more and more attention.This paper mainly discusses and studies the non-local means(NLM)denoising methods.The NLM de-noising method has prominent de-noising ability,simple calculation steps,and easy to improve algorithm,but the algorithm still has its shortcomings.By analyzing the classical NLM de-noising algorithm,we can find that the NLM algorithm calculates the edges and details of the image,because the edge pixel points and the flat pixel points are mixed and calculated.Will make the edge of the gray point value of the pixels tend to be flat,resulting in a smooth phenomenon.Therefore,it is difficult to save the fine structure information in the image.To address this shortcoming,this paper combines edge detection with NLM to remove noise from the image based on the classical NLM algorithm.The innovation of this algorithm lies in that not only the spatial distance but also the edge information of the image is used when weighting pixels.Before performing the non-local mean calculation on the image,edge pixels and flat pixels in the image are detected by edge detection.In the calculation of the subsequent weight distribution and weighted average,the edge pixel points are calculated separately from the flat pixel points.This not only reduces the weight of dissimilar pixels,but also gives greater weight to pixels with similar gray values and structure information.At the same time,the image smoothing of the edge information is reduced.