图像深度估计的光学微分方法

讨论了光学微分方法在图像深度估计问题中的应用。基于线性成像理论对Farid提出的光学微分模型进行了推广，即用于图像深度估计的两幅图像在成像过程中可以满足任意阶的线性微分关系。此模型拓宽了光学微分的概念，使两次成像之间关系有了更多的光学微分形式。围绕如何选择合适的光学微分关系以使系统的整体性能达到最优，分析了光学成像系统的参量对于图像深度估计的精度以及纵向分辨力的影响，并且对光学微分方法中的关键光学元件—光学掩模板的构建方法及优化问题也作了初步的探讨。

The application of optical differentiation method in passive range estimation was discussed. The concept of optical differentiation is extended to a more general model based on the linear property of imaging system. In this model, the distribution of the images' intensity for depth recovery can be formulated as an arbitrary order linearly differential form and it gives us more selections of the optical differential form between the two processes of imaging. For the improvement of the ranging system, the influence of the optical system's parameter on the ranging precision and the longitudinal resolution is analyzed. The design and optimization of the optical masks are also briefly investigated.