Research On Infant Brain MR Image Enhancement And Denoising Algorithm Based On Filtering Method And Probability Model

Magnetic resonance imaging(MRI)is playing an increasingly important role in the current diagnosis and clinical application as a powerful medical imaging technology,its appearance has greatly improved the level of medical diagnosis.However,the brain MR images are blurred,inhomogeneous and noisy.The brain MR images of infants can’t reflect the signal intensity of the lesions accurately and can’t be expressed objectively.The enhancement and denoising of the brain MR images can help doctors to diagnose and subsequent processing such as brain image segmentation.Therefore,the brain image denoising and enhancement have great significance.In this thesis,the enhancement and denoising methods of brain MR images in infants are studied deeply,and the existing problems in the original algorithms are improved to some extent.Specific research work and innovation are as follows:(1)Fractional differential infant brain MR image enhancement algorithm with fuzzy edge detection is studied.When classical fractional differential algorithm enhances brain images,at the same time it also increases the number of acnodes in the image,this problem is also called over-enhancement.The improved fractional differential with fuzzy edge detection algorithm modifies the fixed order number as adaptive number.According to the experimental results,this algorithm not only solves the problem that there are many edge acnodes after enhancing the brain image,at the same time the algorithm reduces the attenuation effect of the traditional fractional differential algorithm for image texture details and smooth region.The experimental statistical tables show that the proposed algorithm has advantages over the traditional algorithms,and the validity of the fusion algorithm is verified.(2)Bilateral filtering of infant brain MR image denoising algorithm with GLCM and median filtering is studied.In view of the problem that the traditional bilateral filtering algorithm is not effective in removing mixed noise,the gray level co-occurrence matrix and median filtering are introduced into the traditional bilateral filtering algorithm.The weights of median filtering are determined by angular second moment that is a statistical characteristic of gray level co-occurrence matrix.The weights of bilateral filtering are larger in the texture changing fast;Conversely,the weights of median filtering are larger in the flat region of image.So,both edge and flat section have better denoising effect.The visual effect of the filtered image and peak signal to noise ratio(PSNR),mean square error(MSE)of the two evaluation indexes prove the effectiveness of the proposed algorithm.This algorithm can effectively improve the shortcomings of the traditional bilateral filtering which can only remove the Gauss noise but can’t remove the salt and pepper noise,the algorithm has achieved good results in removing the mixed noise.(3)Non-local collaborative filtering algorithm for infant brain image denoising based on maximum likelihood estimation.For better removal of Rician noise in the brain MR images,this thesis mainly studies the principle of maximum likelihood estimation method(MLE)and non-local means(NLM)algorithm.The algorithm also fuses K-S test method and the idea of collaborative filtering and forms a collaborative filtering algorithm based on maximum likelihood estimation.The experimental results show that the algorithm has better visual effect on removing Rician noise.Two evaluation indexes of peak signal to noise ratio(PSNR)and mean square error(MSE)verify that the proposed algorithm has obvious advantages compared with the other four algorithms in the experiment.This algorithm can remove the Rician noise in the brain MR images of infants and young children,lay a good foundation for the doctor’s initial diagnosis and other follow-up processing such as segmentation.