基于阵列图像的自适应光场三维重建算法研究  

Study on adaptive light field 3D reconstruction algorithm based on array image

作　　者：丁伟利[1] 陈瑜[1] 马鹏程[1,2] 黄向生[2] 张旭光[1] Ding Weili Chen Yu Ma Pengcheng Huang Xiangsheng Zhang Xuguang(Institute of Electrical Engineering, Yanshan University, Qinhuangdao 066004, China Institute of Automation, Chinese Academy of Sciences, Beijing 100190, China)

机构地区：[1]燕山大学电气工程学院,秦皇岛066004 [2]中科院自动化所,北京100190

出　　处：《仪器仪表学报》2016年第9期2156-2165,共10页Chinese Journal of Scientific Instrument

基　　金：国家自然科学基金(61271409);河北省自然科学基金(F2012203185);燕大青年教师自主研究计划课题(15LGA014)项目资助

摘　　要：针对现有光场图像获取困难,深度重建过程中遮挡以及亮度变化较大区域匹配效果差、稳健性低等问题,提出了基于单反相机的光场图像获取方法以及EPI自适应三维重建算法。在图像预处理阶段,该算法利用双边滤波器对EPI进行去噪,并通过交叉检测模型求得边缘区域。在边缘深度求解以及深度扩散阶段,算法在先验似然策略的基础上,提出EPI自适应框架,通过最大类间方差(OSTU)准则自动设定阈值,舍弃类外点,使距离度量只发生在类内点之间,因此极大地消除了遮挡以及光照变化的影响,提高了边缘深度和内部深度估计的准确性和稳健性。实验结果表明,所提出的系统可以方便地获取阵列图像,成本低、操作方便,且提出的算法能较好地估计场景的深度信息,并实现场景的三维重建,比以往算法在精度上有较大提高。Aiming at the problems ofexisting light field image acquisition difficulty, poor matching effect and low robustness of the regionswith occlusion and large luminance fluctuation in depth reconstruction process, the light field image acquisition method based on SLR camera and adaptive 3D reconstruction algorithm based on EPI are proposed. In the image pre-processing stage, this algorithm utilizes bilateral filter to denoise EPI and obtains the edge region through cross detection model. In the edge depth solving and depth diffusion stage, on the basis of priorilikelihood strategy, the algorithm proposes an adaptive framework of EPI. Thethreshold is automatically set according to the criterion of maximum between class variance, the points outside class are discarded, which makes the distance measurement only take place among the points within class. So, the influence of occlusion and luminance fluctuationis greatly eliminated, and theaccuracy and robustness of edge depth and internal depth estimation are improved. Experiment results show that the proposed system can easily acquire the array image, and features low cost and convenient to operate; the proposed algorithm can estimate the depth information of the scenenicely, realize 3D reconstruction of the scene, and the accuracy of the algorithmis improved greatly compared with that of the existing algorithms.

关 键 词：光场图像 深度重建 自适应EPI 最大类间方差 

分 类 号：TP391.9[自动化与计算机技术—计算机应用技术] TH164[自动化与计算机技术—计算机科学与技术]