带预处理的双树复小波医学超声图像去斑

Medical ultrasound image despeckling based on dual-tree complex wavelet transform with preprocessing

针对现有方法在去除医学超声图像斑点噪声上的局限,提出一种带预处理的双树复小波变换的去斑算法。先从超声I/Q图像中提取实部,并从中"盲估计"出系统的点扩散函数;然后通过维纳滤波,估计组织反射率图像;接着对该图像作同态变换将乘性噪声转化为加性,并进行噪声高斯化处理;最后采用双树复小波变换和双变量收缩函数,获得去斑后的超声图像。分别对同质和边界区域的超声仿真图像、实际超声血流图像采用本文方法和现有方法进行比较实验,结果表明,采用本文方法可将超声图像的斑点信噪比和边界保留指数平均提高2.066和4.091倍,归一化均方差平均降低3.831倍,整体性能指标优于现有方法。

Aiming at the inherent limit of existing methods for medical ultrasound image despeckling,a novel algorithm is proposed based on dual-tree complex wavelet transform（DTCWT） with preprocessing procedure.Firstly,the real part is extracted from the in phase/quadrature（I/Q） ultrasound image,and the point spread function（PSF） is blindly estimated from the real image.Then,the Wiener filtering is implemented to estimate the tissue reflectivity image.Secondly,the homomorphy transform is utilized to turn multiplicative noise into additive one,and the noise is processed with ＂Gaussianization＂.Lastly,DTCWT and bivariate shrinkage function are adopted,and the despeckled image is obtained.Comparison experiments were carried out on homogenous and heterogeneous regions of simulated ultrasound images,and also on in vivo ultrasound blood images.Experiment results show that the proposed method improves the speckle-signal-to-noise ratio（S-SNR） and the edge preservation index（FOM） of the I/Q images by an average factors of 2.066 and 4.091,respectively.Meanwhile,the proposed method also reduces the normalized mean-square error（NMSE） by an average factor of 3.831.The simulation and in vivo results indicate that the proposed method has a better overall performance than existing methods.