摘要
研究了光场数字重聚焦聚焦三维重建新方法,探讨了光场数字重聚焦参数λ的取值以及拍摄条件对重建结果的影响。空间中,每一个点和每一个方向的辐射函数总和就是光场,光场数据包含了空间全部三维信息。借助计算成像技术,光场数据可以重构出序列数字重聚焦图像,实现空间的三维重建。实验时,首先用Lytro光场相机获取空间数据;然后用LytroDesktop软件进行数字重聚焦处理,重构出序列焦平面图像;最后用Halcon软件的聚焦深度(depth-from-focus)函数实现三维重建。实验结果显示,参数λ表征了重构平面的深度信息,λ的性质显著地影响重建结果,λ数量多于8、正负λ数目相当时,三维重建结果好。拍摄时,设置的相机参数、目标的反射及场景背景等因素都会对重建结果产生影响。适合参数下,本文方法能够正确重构不同深度结构的目标,重构结果可以任意视角查看。本文方法利用单视图重构设备,实现多视图三维重建结果,拍摄装置小巧,拍摄过程快捷,系统稳定性好,降低了拍摄难度和重建算法的复杂度,拓展了DFF算法的适用深度范围,适合大景深场景和运动目标的三维重建。
A new method for 3D reconstruction based on light field digital refocusing and depth-from-fo- cus （DFF） is presented. The effects of the value of digital refocusing parameter λ as well as shooting conditions on reconstruction results are discussed. Light field is the sum of the radiation functions from all points and directions in the space, which means that the light field data includes all the 3D geometric information. Series of refocusing images at different focal planes could be acquired by calculating these light field data, which could retaliate 3D reconstruction via calculation imaging. During the experiment, first, spatial information is obtained by light field camera-Lytro. Then, a succession of clear images at dif- ferent focal planes are reconstructed by digital refocusing with Lytro Desktop software. At last, 3D re- constructed images are achieved by depth-from-focus function with Halcon machine vision software. The parameter ）, characterizes the depth information of refocusing plane, which would impact the reconstruc- tion greatly. Reconstructed results show that the accuracy of reconstruction would be higher with more quantities--at least 8, and higher symmetry in positive and negative values. At the same time, the recon- struction result would also he affected by some other factors, such as the settled camera parameters, re-flection and background. When parameters are in the appropriate range,objects in different depths could be reconstructed correctly and the reconstructed image could be observed from any angle. 3D reconstruc tion in multi-view is obtained by the reconstruction equipment in single view, which is portable and small. The photographing process is convenient, and the system is of high stability. It reduces the difficul- ty of photographing and the complexity of reconstructive arithmetic. Besides, using this new method, the depth-from-focus algorithm could be extended to the large-scale scene and moving objects.
出处
《光电子．激光》
EI
CAS
CSCD
北大核心
2015年第5期986-991,共6页
Journal of Optoelectronics·Laser
基金
国家自然科学基金(61307019
11404290)
河南省科技厅重点科技攻关(132102210396)
河南省教育厅科学技术重点研究(13B140395
12A140014
14A140005)
郑州大学创新创业(2013xjxm013
2014xjxm228
2014xjxm206)资助项目
