| Medical image is an indispensable and important reference in clinical auxiliary diagnosis.Medical CT,MR and B-mode ultrasound images are widely used in both routine examination and disease diagnosis.Obtaining clear amd reliable medical images is the key to noise reduction.This paper chooses the above three kinds of medical images as the research objects,aiming at obtaining real texture and clear details of denoising image,proposes several improved denoising methods and combines subjective observation with objective indicators,comprehensively analyzes the denoising effect of various methods.Specific research contents are as follows:1.Comparing the processing effects of several common noise estimation methods and four classical denoising algorithms in order to provide ideas for the improvement of this algorithm.2.The non-local mean filtering(NLM)method is studied:(1)An improved similarity weight with rotation invariance is proposed to measure the similarity between pixel blocks.(2)An improved noise estimation method based on Laplacian operator is proposed to estimate the noise variance of noisy images.Experiments show that the improved noise estimation method based on Laplacian operator has higher estimation accuracy.The improved NLM method combines the advantages of the classical denoising algorithm and gives full play to the advantages of the original NLM method.The PSNR can be improved by up to 1.5 db.The improved NLM method has more advantages than the original method in terms of visual effect,noise removal,structural similarity,edge detail protection and so on.It is more suitable for processing the above medical image noise.3.The bilateral filtering denoising method is studied,and a new gray similarity measure factor for processing CT,MR and B-mode ultrasound images is proposed.Experiments show that the PSNR of the improved method can be increased by more than 4 db,while SSIM and EPI are closer to 1.This shows that the improved bilateral filtering method can obtain better denoising images,and can effectively reduce the occurrence of over-enhancement while preserving the edges of details.4.The wavelet denoising method is studied:(1)An improved wavelet soft threshold function is proposed.Based on the theory of wavelet transform,the soft threshold function suitable for medical image denoising is selected as the improvement object,and the wavelet decomposition level is added as the adjustment parameter.The experimental results show that the PSNR of this method can be increased by about 3 dB at most,which is better than the original soft threshold method.(2)Non-local mean filtering and bilateral filtering based on wavelet domain are proposed.This method uses the advantage of wavelet to separate signals.Firstly,the high-frequency and low-frequency regions of the image are obtained by wavelet decomposition.Then the improved NLM filtering is used to denoise the high-frequency regions,and the improved bilateral filtering is used to denoise the low-frequency regions.Finally,the denoised images are obtained by signal reconstruction.In the same simulation software environment,the actual effect of the four improved noise reduction methods is compared.Experiments show that:(1)The non-local mean filtering and bilateral filtering based on wavelet domain have the advantages of both NLM and bilateral filtering.Noise points are eliminated to a certain extent,and image edge contour is enhanced moderately.(2)The visual effect of the improved NLM method is the closest to that of the ideal noise-free image,but the visual effect of processing B-mode ultrasound image with smooth texture structure is slightly lower than that of the improved bilateral filtering method.(3)The maximum EPI of the improved bilateral filtering method can be increased to 3,and the image edge details can be enhanced to the greatest extent.(4)The improved wavelet soft threshold denoising method is lower than the other three methods in terms of noise removal and detail protection.Figure 83 Table 21 Reference 82... |
PDF Full Text Request