A Research On The Noise Suppression Algorithm For Medical Ultrasonic Images

Medical ultrasonic images have a wide range of applications in medical diagnostics.However,the application of medical ultrasound images is limited because of the low resolution and severe speckle noise of them.Therefore,remove the speckle noise of the medical ultrasonic image is a research direction with great significance and value.This work mainly studies the de-noising algorithm of medical ultrasound images.The specific works are as follows:1.The frequency domanial layered noise suppression method is proposed.In this work,the soft threshold de-noising method based on non-subsampled contourlet transform(NSCT)has been improved,and a frequency domanial layered image de-noising method has been proposed.In this method,the NSCT-decomposed layers of image are divided into lowfrequency layers,intermediate-frequency layers,and high-frequency layers to be de-noised by separate methods respectively.Among them,the soft threshold de-noising method has been implemented to the intermediate-frequency layers;as for the de-noising of the low-frequency layers and the high-frequency layers,this work proposed a total variational de-noising method based on median filter and phase consistency,as well as an image noise removal method based on the directional local intuitionistic fuzzy entropy,respectively.The specific work is as follows:A.A total variational de-noising method based on median filter and phase consistency has been proposedIn this method,the total variation de-noising method has been improved.Firstly,a control factor which employs the maximum phase consistency moment is proposed to further adjust the diffusion speed and the scale factor of the image in the total variational method.Secondly,a median fusion filter with phase consistency constraints has been proposed and introduced into the total variational de-noising method.In this filter,the phase consistency information of the image is used to control the proportion of images before and after the median filtering process in the fusion step to suppress the noise and maintain the edge and texture information of the image.B.An image noise removal method based on the directional local intuitionistic fuzzy entropyhas been put forwardIn order to improve the de-noising ability of NSCT decomposed high-frequency coefficients,an image de-noising method based on directional local intuitionistic fuzzy entropy has been put forward.In this method,we utilized the rectangle direction sampling window in the local intuitionistic fuzzy entropy method based on image gradient information to improve the recognition ability of useful signals in high frequency coefficients.Meanwhile,the directional local intuitionistic fuzzy entropy of the obtained image is employed to de-noise the coefficients of the high-frequency layer.Experiential results show that this method can improve the de-noising effect of the high-frequency layers,and can also better protect the high-frequency information of the image.2.An image multi-layer filtering,fusion,and enhancement method based on block matching three-dimensional(BM3D)collaborative filtering method has been advancedIn order to improve the de-noising ability of BM3 D method for speckle noise,in this paper,the NSCT hard threshold de-noising method and BM3 D method are combined,and an image multi-layer filtering,fusion and enhancement method is proposed.Firstly,we designed the NSCT hard threshold de-noising and enhancement method to pre-process the noisy images.Secondly,the basic estimations of images before and after the NSCT hard threshold de-noising and enhancement were calculated.Thirdly,an image multi-layer fusion and enhancement method in the NSCT transform domain has been proposed to merge each corresponding layer of the basic estimation results above,and the highfrequency layers were enhanced.Finally,a basic estimation is performed on the NSCT reconstructed result of step 3.This basic estimation result is used to guide the final estimation of the original noisy image.Experiments show that this image multi-layer filtering,fusion and enhancement method can remove the speckle noise in the image,and protect the high frequency information of the image as well.