Edge-guided TV p regularization for diffuse optical tomography based on radiative transport equation
Edge-guided TV p regularization for diffuse optical tomography based on radiative transport equation
复制标题
基于辐射传输方程的漫射光学层析成像边缘引导TV p正则化
DOI：
10.1088/1361-6420/aadb23
复制
发表时间：
2018
期刊：
影响因子：
2.1
通讯作者：
Jinping Tang
中科院分区：
文献类型：
--
作者：
Shanshan Tong;Bo Han;Jinping Tang
在光学层析成像中，将TVp最小化方法与边缘引导策略相结合，引入边缘引导TVp正则化方法，同时恢复散射和吸收系数。TVp最小化方案由数据保真度项和基本光学系数梯度的-范数组成。为了使最小化问题数值上易于处理，Huber函数被用来局部光滑的范数，以获得一个微分目标函数。滞后扩散牛顿迭代法被用来寻找上述Huberized目标函数的最小值。边引导策略以加权矩阵的形式插入到每次迭代中。此外，归一化技术被纳入我们的算法，以减少串扰。与正则化方法相比，该方法在保持目标形状和大小以及去除背景起伏方面具有明显的优势。然后，所提出的方法被施加到图像的组织中存在的非散射或低散射层。结果表明，我们的方法是能够成像的目标，在组织中包含一个非散射层使用减少测量数据。此外，它持有的承诺，共同成像的目标和低散射层在一定的噪声水平。
An edge-guided TVp regularization is introduced to recover scattering and absorption coefficients simultaneously in optical tomography, which combines the TVp minimizing scheme with an edge-guided strategy. The TVp minimizing scheme consists of a data fidelity term and an -norm of the gradients of underlying optical coefficients. To make the minimization problem numerically tractable, the Huber function is utilized to locally smooth the -norm to obtain a differential objective function. The lagged diffusivity-Newton iteration is applied to find the minimizers of the above Huberized objective function. An edge-guided strategy is inserted into each iteration in the form of a weighted matrix. In addition, a normalizing technique is incorporated into our algorithm to reduce the cross-talk. Compared with regularization, the proposed edge-guided TVp regularization presents superiorities on keeping the shape as well as the size of targets and removing background undulations. Then the proposed method is applied to image the tissue in the presence of a non-scattering or a low-scattering layer. It is shown that our method is capable of imaging the targets in the tissue containing a non-scattering layer using reduced measurements data. Moreover, it holds promise for jointly imaging both targets and the low-scattering layer under a certain noise level.