基于平移不变多小波方法的磁共振去噪研究

Study of MRI Denoising Based on TI Multiwavelet Thresholding

为了提高影像的质量,将一种新颖的基于平移不变原理的多小波方法引入到磁共振的复数信号去噪中.通过对多小波向量系数在一定范围内的循环平移求平均估计子,获得一个改进的阈值估计子,并在不同尺度的高频子带内取双元硬阈值以实现最小化期望风险.实验结果显示,该方法相对于目前几类高效算法和传统的平移不变单小波方法,在消除噪音和抑制伪影的同时能更好地保留边缘特征,并显著改善了信噪比和均方根误差.

A novel multiwavelet thresholding based on translation-invariant theory is introduced to complex denoising of MR data to advance the image quality. An improved thresholding estimator is calculated by averaging estimators for a range of eireulant shifts of sealed vector eoeffieients, and bivariate hard thresholding is adopted in the high frequency sub-bands at each scale to minimize the expected risk. Experiments show that the proposed scheme better preserves edges while suppressing noise and artifacts than some of recent effective methods and traditional TI scalar wavelet thresholding; moreover, both the SNR and RMSE are improved obviously.