Improvement Of BM3D Algorithm Based On Edge Detection And Image Clustering

The Block-matching and 3D filtering(BM3D)algorithm absorbs ideas from various classical filtering algorithms to achieve excellent performance.However,its high time cost,low noise reduction efficiency,and inability to preserve texture edge information while denoising make it difficult to apply this algorithm in practice.To improve the problems that exist in the BM3 D algorithm,corresponding optimizations have been made to both its basic estimation stage and final estimation stage.The following are the optimization ideas and research work of this paper.(1)The reason why the BM3 D image denoising algorithm has such outstanding denoising performance is due to the fact that the algorithm combines the idea of non-local mean filtering algorithm.That is,the algorithm searches for similar blocks in the global range,but this also increases the time cost of the algorithm.In order to improve the efficiency of the algorithm,this paper optimizes the search range of similar blocks:before the final estimation stage,an improved K-Means clustering algorithm is used to divide the image into several similar regions and constrain the search range of similar blocks within its own similar region,which solves the problem of blind search of the algorithm in the image.Since irregular graphic blocks will appear in the same region,an adaptive matching algorithm for irregular graphic blocks is proposed to reduce interference from redundant information.Finally,through experimental simulation results analysis,this paper’s improvement scheme can effectively improve the efficiency of the algorithm.(2)The BM3 D image denoising algorithm simply moves the search window in horizontal and vertical directions when searching for similar blocks,ignoring the edge information in the image.This leads to the edge blocks being improperly assigned to the three-dimensional array,resulting in the loss of edge information during the filtering process.To address this issue,this paper proposes an adaptive Canny edge detection algorithm based on morphological filtering and Otsu algorithm,which identifies the edge areas in the image.The block matching process is then carried out along the direction of the edge area to ensure that edge image blocks are assigned to the correct threedimensional array,thereby improving the denoising and preservation of edge information.According to the experimental results,the improved algorithm can simultaneously remove noise and preserve edge information.(3)The above improvement strategies have improved the performance of the BM3 D algorithm to a certain extent,but they also have their own limitations.Therefore,this paper combines the advantages of both to propose a new image denoising algorithm to compensate for the shortcomings of the two algorithms.The proposed algorithm is applied to denoise lung CT images,and its feasibility is confirmed by extensive experimental data.