光学层析成像是一个病态重建问题,为克服重建过程的病态性,提出将多准则优化理论引入到图像重建中。利用了三个用于光学层析图像重建的准则：平方误差函数、图像熵函数和局部平滑函数。采用向量优化方法将多准则优化问题转化为单准则优化问题求解。为了确定各个目标函数间的权重系数,提出一种动态权重系数求解方法。重建过程目标函数关于光学参量的梯度计算是关键,因此提出一种基于梯度树的计算方法。实验过程中对多准则重建结果和基于平方误差函数的单准则重建结果做了比较,证明该方法能够克服传统的偏重单一目标的单准则重建的不足,有效地重建光学层析图像,提高图像重建质量。

The reconstruction of optical tomography from measurement data is an illposed problem. For such an illposed problem, a multicriterion strategy is proposed, in which three criteria of squared error function, image entropy and local smoothness function are used for the reconstruction. The multicriterion reconstruction problem is transformed to a single criterion problem by using the vector optimization method. A dynamic weight coefficient solution is proposed to determine the weight coefficient of objective function. For the realization of optical tomography reconstruction, a gradient tree based algorithm is proposed for the gradient computation of the objective function with respect to optical parameters. Different results from the multicriterion reconstruction and single criterion reconstruction based on the squared error function are presented and compared. Experimental results show that this overcomes the shortcoming of conventional single creterion reconstruction only depending on the single objective and is valid for optical tomography reconstruction and the image quality is significantly improved.