融合多种先验信息的光学层析图像重建

Optical Tomography Reconstruction with Multiple Priori Information

为克服光学层析成像的病态特性,将三种不同的先验信息模型引入到基于辐射传输方程的光学层析成像中,给出了对应于三种重建模型的重建算法,其中在线过程作为先验信息的重建算法中,提出一种神经网络优化算法。最后对这三种先验信息模型进行了综合比较和讨论,并在图像的平滑性、边缘保留特性、算法复杂性及重建效果方面对其进行定量比较,提出一种分析平滑性的算子和一种基于梯度的衡量边缘保留特性的算子。实验证明基于三种先验信息模型的重建算法均能提高重建图像质量,其中,平滑先验模型的平滑效果最好,马尔可夫随机场模型有较好的边缘保留特性,线过程模型在平滑图像的同时可以更好地保留图像边缘,并且重建精度较高。

Three kinds of priori information models are imported into the optical tomography based on the equation of radiative transfer in order to overcome ill-posedness. Specific reconstruction algorithms corresponding to them are given. A neural network optimization approach is presented when the line process is used. Comparison and analysis on reconstructed results with different priori items are given. In order to analyze the effect of prior information models on reconstructed images, quantitative comparison are conducted in terms of smoothness index, edge-preserved feature and algorithm complexity. A smoothness operator and a mathematical model based on the gradient for measuring edge-preserved feature are proposed. Experimental results show that the quality of reconstructed images can be improved by these priori items. It can be concluded that the smoothness of smoothing model is best; Markov random field has a better edge-preserved feature; The line process model can preserve better image edge while smoothing the image and achieving higher reconstruction accuracy.