计算摄像中多视场光场高精度成像方法研究

项目名称： 计算摄像中多视场光场高精度成像方法研究

项目编号： No.61471213

项目类型： 面上项目

立项/批准年度： 2015

项目学科： 无线电电子学、电信技术

项目作者： 王兴政

作者单位： 清华大学

项目金额： 85万元

中文摘要： 光场研究，作为计算摄像学的核心研究领域，在航天、军事、医疗、教育、影视、公共安全等领域有着广泛的应用。针对现有光场成像技术面临的视场单一、建模精度不高、成像分辨率低下等问题，本项目着眼于计算摄像学中多视场光场高精度成像这一热点问题展开。提出多棱面环形反射镜＋微透镜阵列进行多视场光场采集的成像方法，通过对光路和光学元件进行优化设计，实现高质量的多视场光场采集；同时，基于多视场光场采集原理，建立目标景物深度估计的计算模型，提出完整的多视场深度信息提取方案，并基于深度信息，开展数字重聚焦算法和对象三维建模的研究。最后，提出基于傅里叶切片理论的光场超分辨率方法，解决光场成像空间-角度分辨率之间的矛盾，实现高精度的光场图像重建，从而提高了对成像目标的重建精度，对多维景物的采集、重构和内容生成机理等方面的研究也起到重要的作用。

中文关键词： 光场成像；多视场；深度估计；光场超分辨率

英文摘要： Light field research, as the core research area of computational photography, which has a wide range of applications such as in aerospace , military, medical , education, film and television and public security. Aiming to solve the problem lie in the current research of light field imaging, i.e., single field of view, low accuracy of 3D-modeling and poor resolution of light field images, this project focuses on the study on high-accuracy light-field imaging system with multi-field of view. Firstly, we utilize an annular mirror and micro lens array as the main optical component. After optimally design of the optical path and parameters, light filed data with multi-field of view would be captured simultaneously. Secondly, based on the principle of light field imaging, the computational model for object depth estimation is established. Furthermore, based on the obtained depth map, digital re-focusing and object 3-D modeling algorithms will be developed. Finally, we propose a light field super-resolution algorithm based on Fourier Slice Theorem. By this approach, the confliction between the spatial and angular resolution would be dissolved, and high-resolution of image reconstruction would be achieved.

英文关键词： Light field imaging;Multi field of view;Depth estimation;Light field superresolution