DT-MRI张量值图像的一种加权方向距离滤波方法

为了正则化扩散张量磁共振成像所产生的扩散张量场,提出了一种非线性张量值加权方向距离滤波方法.首先通过张量间的二次点积定义了张量值图像的方向滤波,然后结合张量中值滤波引入了一个联合最小化准则,在此准则的基础上给出张量值方向距离滤波器的一般结构,并使用权重矢量在距离和方向域上控制滤波器输出,获得更加稳健的处理结果.在合成二阶张量场和真实扩散张量数据集上进行了实验.客观结果比较和滤波图像的主观评估表明:该滤波方法可在保持张量场细节不变的同时有效地消除噪声,是磁共振扩散张量场的一种有价值的滤波工具.

Weighted directional-distance filtering method for DT-MRI tensor-valued images

A nonlinear tensor-valued weighted directional-distance filtering method is proposed for regularizing the diffusion tensor field produced by diffusion tensor magnetic resonance imaging.First,the directional filter for tensor-valued image is defined through double dot product of tensors,then the tensorvalued median filter is combined to introduce a joint minimization criterion.Based on this criterion,the generalized tensor-valued directional-distance filtering structure is given.Filter output can also be controlled by employing weights vector on distance and direction domains,more robust results are acquired consequently.Experiments were performed on both synthetic second-order tensor field and real diffusion tensor data set.The comparison of objective results and subjective(evaluation) of the filtered images indicate that the proposed filtering method can effectively remove noise and keep tensor field details unchanged.It may be a promising filtering tool for the magnetic resonance diffusion tensor field.