Research And Platform Construction Based On Affine Non-local Mean Denoising Algorithm

In real life,unless an image is completely generated by a computer,it will definitely be affected by the acquisition noise,resulting in a decline in image quality.Therefore,image denoising has become one of the key research topics in the field of image processing.Researchers expect to use the image denoising technology to maximize the image quality,and preserve the structure and features of the image while removing noise.As one of the classic image denoising algorithms,non-local mean has always been a research hotspot.For its shortcomings and deficiencies,various improved algorithms about it are constantly being proposed.In order to solve the problem that regular square image blocks are easy to produce noise on the denoising edge,the affine non-local mean denoising algorithm proposes to use affine co variant ellipse regions instead of square image blocks,and introduces an affine invariant similarity measure to obtain a richer block space greatly improves the denoising effect.This paper conducts in-depth analysis and research on the affine non-local mean denoising algorithm,and proposes further improvement methods for its shortcomings.The specific research work is as follows:Firstly,propose a fast affine non-local mean denoising algorithm.Aiming at the problem that the affine non-local mean algorithm takes too long in the denoising process,an acceleration algorithm is proposed for it.The algorithm first uses the included angle of the feature vector of the affine covariant structure tensor to replace the main direction of the SIFT operator,and then uses the fast Fourier transform to accelerate the affine in the affine invariant similarity measure in the affine non-local mean method Calculation of similarity measures between covariant feature regions.Experiments show that compared with the original affine non-local mean algorithm,the speed has been greatly improved.Secondly,propose an affine non-local mean denoising algorithm based on parameter adaptation.Inspired by the affine non-local mean denoising algorithm,in order to improve the effect of edge denoising,it is proposed to change the shape of the search window and image block,and generate stable elliptical search blocks and image blocks according to the internal structure of the image itself,and adaptively Choose its size;from the perspective of distinguishing texture,the Canny edge detection operator is used to realize the adaptability of the smoothing coefficient.The denoising effect is further improved.In addition,a visual experience platform was built around the affine non-local mean algorithm.The platform includes two functions:image denoising and algorithm experience.Users can not only use the one-click denoising function,but also experience non-local mean,fast non-local mean,affine non-local mean,and fast affine non-local mean according to their needs.And the five function modules of parameter adaptive affine non-local mean,to understand the image denoising effect of each algorithm under different parameters.Through this platform,users can continuously deepen their understanding of algorithm principles in dynamic selection and experience.