Research On Magnetic Resonance Image Segmentation Based On Dual-tree Complex Wavelet Transform

In recent years,with the rapid development of modern medical image technology,the accurate segmentation of medical images is significant for doctors to diagnose and analyze the etiology.Magnetic Resonance(MR)imaging is the use of magnetic resonance phenomena,and the electromagnetic wave signals are processed by computer to obtain a tomographic image.It has the characteristics of arbitrary slice imaging and high soft tissue resolution,and is suitable for the examination of different diseases of various systems in the whole body.However,due to the influence of external factors and the defects of the imaging device,MR images have problems such as boundary blur,intensity unevenness and random noise,which poses a great challenge for accurately segmenting MR images.In order to solve this problem,this paper effectively combines the dual-tree complex wavelet transform with other methods,and proposes a variety of MR image segmentation methods.The main work of this article is as follows:(1)The MR image of the mammary gland has the characteristics of uneven gray scale,complicated noise and difficult to segment,a segmentation method based on dual-tree complex wavelet transform and density clustering is proposed.Firstly,the image is denoised by using complex wavelet domain bivariate model combined with anisotropic diffusion function,eliminating the impact of noise on segmentation result;Then simple linear iterative clustering(SLIC)algorithm is used to obtain the neighbors of each super pixel,thereby reducing the time of searching for the nearest neighbor of each sample in KNN-DPC algorithm.Finally,nearest neighbor sample density information of super pixel region is introduced,and distribution strategies from KNN-DPC algorithm are used for adaptive clustering.The segmentation results of simulation and clinical data show that the proposed algorithm can segment breast MR images effectively.(2)Aiming at the poor selectivity of multi-resolution Markov field(MRMRF)model,a multi-scale transform is proposed by using non-subsampled double-tree complex wavelet(UDTCWT)transform instead of wavelet transform,and the improved k-means algorithm is used to obtain coarse scale segmentation results.Then,the coarse-scale segmentation results are segmented by the MRMRF model.Aiming at the problem of edge segmentation blur caused by gray level inhomogeneity of MR images,it is proposed to introduce variable weight parameters in MAP criterion estimation,and then optimize the segmentation effect.(3)Considering the large deviation of KFCM algorithm in MR image segmentation,the nonsubsampled dual-tree complex wavelet transform is used to pre-process the segmented image,and then different weight coefficients are given to the high and low frequency components obtained from the decomposition to reconstruct the function to be segmented for KFCM.The neighborhood spatial information is used to design the spatial function to improve the objective function and effectively solve the isolated region.The correct classification problem enables the algorithm to segment MR images with higher accuracy.