基于梯度树的光学层析正则化重建

光学层析成像是一个病态重建过程，为降低重建过程中的病态特性，需加入合适的先验信息。目前，大多数重建都是基于扩散方程的，在某些情况下，这种重建会失败。直接基于玻耳兹曼传输模型，并以图像熵为正则化项的梯度迭代重建是一种有效的方法。该方法中，梯度计算是个难点。对此，提出一种基于梯度树的求解方法，降低光学层析图像重建的病态性，有效地重建光学层析图像。

Regularized reconstruction for optical tomography based on gradient tree

It is well known that optical tomography(OT) is an ill-posed problem and some proper a priori information is incorporated in order to decrease the ill-poseness.At present,most of the reconstructions are based on diffusion equation,which will fail in some cases.Hence,the reconstruction process is put forward based on Boltzmann transport model directly with the image entropy as the re-gularized item,which is implemented by the gradient-based iterative reconstruction scheme,but the gradient computation of objective function with respect to opticalparameters is difficult.Soa gradientcalculation strategy based on gradient tree is proposed.Experimental results show that OT image is reconstructed effectively,its ill-poseness is decreased,and the reconstruction quality at the same time is improved.