Research On MR Image Denoising Algorithm Based On Genetic Algorithm Optimization Of Multi-path Matching Pursuit

In recent years,medical image processing has played an important role in improving the accuracy and effectiveness of modern medical diagnosis.Magnetic Resonance Imaging(MRI)makes MR images the first choice for assisted medical diagnosis because of its advantages in sensitive tissue,clear imaging,and distinct differentiation of tissues and organs.However,in some cases,when the MRI signal is very weak due to a specific acquisition sequence,the MR image is affected by artifacts and random noise,resulting in lower image quality.Noise becomes a critical issue that to be properly handled.A tradeoff between noise reduction and the preservation of actual detail features has to be made in the way that enhances the diagnostically relevant image content.Therefore,noise reduction is still a difficult task.In this paper,the multipath matching pursuit algorithm is firstly applied to MR image denoising,and then a multipath matching pursuit algorithm optimized by genetic algorithm is proposed.Then the genetic algorithm is further improved and finally applied to MR image denoising.The research works in this paper include:1)The multipath matching pursuit is used for MR image denoising.In the algorithm,the Gabor function is used to generate an over-complete dictionary.The effects of different atomic numbers and different scale atoms on the performance of the algorithm are discussed.It effectively combines the advantages of local representation characteristics and direction selectivity of the Gabor atom in dictionary and multipath matching pursuit algorithm.It not only overcomes the shortcomings of the single candidate set of the original algorithm,but also improves the denoising performance.In the reconstruction experiment of standard test image and simulated MR image,which demonstrates good performance in image detail feature retention.2)An improved multipath matching pursuit algorithm for image reconstruction to achieve denoising purposes.The algorithm introduces adaptive genetic algorithm optimization in the process of solving multiple candidate atoms matching the local image features in each iteration of the multipath matching pursuit algorithm,which combines the advantages of adaptive genetic algorithm and multipath matching pursuit algorithm.It not only avoids the genetic algorithm's easy to fall into the local optimal defect,but also obtains the best matching parameters with higher precision,and effectively reduces the computational complexity of the multipath matching pursuit algorithm,and overcome the shortcomings of the multipath matching pursuit algorithm because the calculation is too large to promote the application.In the reconstruction experiment of standard test image and simulated MR image,the algorithm shows good denoising effect,and the reconstructed image precision and reconstruction time are obviously improved.Finally,the algorithm is applied to the denoising of clinical MR images.The experimental results show the clinical application value of the algorithm.3)Since the adaptive genetic algorithm will fall into the possibility of local optimization.Furthermore,the self-identification crossover operator is used to improve the convergence speed,and the self-adaptive mutation operator is used to increase the population diversity.The effective combination of the two operators makes the genetic algorithm not only retain good individuals in each iteration process,but also reduce the ability of the algorithm to fall into local optimum.The standard test function experiments show the optimization ability and evolutionary stability of the algorithm.Finally,the algorithm is introduced into the iterative optimization of multipath matching pursuit algorithm for image denoising.The algorithm effectively improves the optimality of the selected atomic parameters and shows a good denoising effect.