Speckle Filtering Of Medical Ultrasonic Images Based On Translation Invariance In Shearlet Transform

In medical clinical ultrasound diagnosis,the using of ultrasonic diagnosis technology has became more and more important.However,the presence of speckle noise in the process of ultrasonic imaging,reduces the image quality and bring inconvenience to the doctor's diagnosis.The study found that the nature of medical ultrasound image denoising is using noise image to estimate free image which can reflect the characteristics of human body organs and get the optimum estimate of real images and it brings more important practical significance for diagnostic accuracy.In order to suppress the speckle noise in ultrasonic images,according to the uniqueness and the statistical features of the speckle noise in medical ultrasound images,a new denoising algorithm is proposed in this paper,which named "Speckle filtering of medical ultrasonic images based on translation invariance in shearlet transform".The shearlet coefficients in high frequency sub-band will be shrunk and the shearlet coefficients in low frequency sub-band will be filtered by specific filter.Firstly,the non-subsampled Laplacian pyramid filter is used to decompose the noise image into multi-scale decomposition,include high-frequency sub-band and low-frequency sub-band.Then the non-subsampled direction filter bank is used to decompose the multidirection of the obtained high-frequency sub-band.Secondly,combined with the correlation between shearlet coefficients in transform domain,an improved threshold shrinkage function based on the traditional threshold functions shrink the shearlet coefficients in high frequency sub-band,then the shearlet coefficients in low frequency is filtered by guide filter.Last,The filtered image is then obtained by using the inverse shearlet transformation.Experiments with the comparison of the other six de-speckling filters are conducted.The results show that the proposed method not only has a strong de-speckling ability,but also keeps the image details,and in the meantime,it doesn't need too much time and thus can meet the real-time requirement of medical ultrasound imaging system.In this paper,the main work and achievements are shown as follows:(1)According to the characteristics and statistical properties of speckle noise indifferent sub-bands of transform domain,an improved threshold function based on the universal threshold function is developed,which determined by the decomposed layers;(2)In this paper,the sheatlet transform solves the limitation of the direction of wavelet transform.In order to solve the ringing phenomenon,put forward the non-subsampled Laplacian pyramid filter and the non-subsampled direction filter;(3)Because the shearlet coefficients in the transform domain has a very strong correlation,so a new three shrinkage algorithm based on the MAP shrinkage algorithm is proposed;(4)The shearlet coefficients of low frequency sub-band will be filtered by the specific filter,which can balance the denoising and preserving the edge information,this paper also give the obtained method of the guide figure when using the guided filter;(5)In the simulation experiment and medical ultrasound images experiment,some evaluation indexes and analysis methods are used to analysis the results of contrast experiment compared with other filters.