Research On Sub-modal Denoising Method Of Medical Ultrasound Image Based On Genetic Algorithm To Optimize 2D-VMD

Medical ultrasound imaging has been widely used in clinical medical diagnosis due to its non-invasive,real-time,non-radioactive,and low-cost advantages.However,in the process of medical ultrasound imaging,a unique speckle noise will be generated,which will seriously reduce the visibility of the image and increase the difficulty of medical clinical diagnosis.Therefore,it is necessary to denoise the noisy image.Traditional image denoising algorithms may have problems such as blurred image edge details and poor denoising effect after denoising.Based on this,this thesis proposes a hybrid denoising method that optimizes transform domain sub-modal joint denoising in spatial domain.Firstly,the two parameters of the Two-Dimensional Variational Mode Decomposition(2D-VMD)algorithm are adaptively optimized at the same time by using the genetic algorithm,and then the optimized 2D-VMD is used to decompose the noise image,and with the help of correlation.The coefficients are used to screen out the valid sub-modes,and then FNLM filtering is used to denoised the retained valid sub-modes,and finally the denoised sub-modes are reconstructed to complete the denoising.The specific research content of this article is as follows:1.This thesis decided to use a smooth noise algorithm based on transform domain,2D-VMD algorithm.Using the fully adaptive and non-recursive nature of the algorithm,the noisy image can be decomposed into multiple sub-modes with different center frequencies.Usually,the lowfrequency sub-mode contains a large amount of detailed information in the original image and is doped with a small amount of noise.The sub-modes contain a lot of noise information and carry a small amount of edge information of the original image.Next,use the correlation coefficient to filter out the invalid sub-modes,and then filter and denoise all the valid sub-modes once.It can achieve the purpose of removing high-frequency noise of the image and retaining its own lowfrequency characteristics.2.The 2D-VMD algorithm needs to set two important parameters before running.Whether the two parameters are suitable or not will have a direct impact on the image decomposition and even the results of subsequent denoising.Manual selection of parameters based on experience will not only greatly reduce the performance of the algorithm.Efficiency,and it is still impossible to determine whether the result of the 2D-VMD algorithm running at this value is the best decomposition effect.Therefore,in order to solve the problem of parameter selection,this thesis considers using genetic algorithm to adaptively optimize two important parameter values of 2DVMD algorithm at the same time,and verifies the proposed method through experiments.3.After conducting a large number of experiments on experimental digital images and real ultrasound images provided by hospitals,and using various quality evaluation indicators to compare and analyze other five commonly used speckle denoising algorithms,the experimental results show that the method proposed in this thesis is better.Other comparison algorithms have better denoising effects and the ability to retain image texture details,and the objective quality evaluation indicators have also been significantly improved.4.In order to simplify the whole denoising process,it is decided to link the algorithm in this thesis to the MATLAB GUI medical ultrasound image processing platform built by our research group to further increase the practical richness of the platform and improve the efficiency of denoising processing.