改进的LLMMSE滤波器对扩散加权图像的降噪

Improved LLMMSE filter in denoising of diffusion weighted image

扩散加权图像具有多边界的特点,在扩散加权图像中,准确的边界信号对扩散张量图像的计算尤其重要。通过对局部线性最小均方误差滤波器(Local Linear Minimum Mean Square Errorfilter,LLMMSE filter)在图像边界处降噪特点进行分析,提出基于最小方差数据集的改进的LLMMSE滤波算法。通过将所提算法应用于模拟数据及真实数据,以及与LLMMSE算法进行比较,验证了本算法具有更好的边界信号降噪能力。

Diffusion weighted images are characterized by multi-boundaries.Accurate boundary signals are essential in comput- ing diffusion tensor images based on DWIs.Through the detailed analysis of the effect of the LLMMSE filter applied in de- noising the image boundary, an improved LLMMSE filter based on the theory of local minimum variance is proposed.After the application of this proposed method in both simulate data and real data and the comparison of the improved method with LLMMSE,this proposed method is proved to produce better effect of denoising the boundary signals.