摘要
光场相机通过单次拍摄可获取立体空间中的4维光场数据,利用光场的多视角特性可从中提取全光场图像的深度信息.然而,现有深度估计方法很少考虑场景中存在遮挡的情况,当场景中有遮挡时,提取深度信息的精度会明显降低.对此,提出一种新的基于多线索融合的光场图像深度提取方法以获取高精度的深度信息.首先分别利用自适应散焦算法和自适应匹配算法提取场景的深度信息;然后用峰值比作为置信以加权融合两种算法获取的深度;最后,用具有结构一致性的交互结构联合滤波器对融合深度图进行滤波,得到高精度深度图.合成数据集和真实数据集的实验结果表明,与其他先进算法相比,所提出的算法获取的深度图精度更高、噪声更少、图像边缘保持效果更好.
The light field camera can obtain the four-dimensional light field data from stereoscopic space with one shot.After that, the depth information can be extracted by multiview characteristic of the light field. However, the existing depth estimation method rarely considers the presence of occlusion in the scene. The accuracy of the extracted depth information is significantly reduced, when the scene is blocked. Aiming at this problem, this paper presents a new depth estimation method of light field image based on multi-clues fusion to obtain high-precision depth information. Firstly, the adaptive defocus algorithm and the adaptive matching algorithm are used to extract the depth information of the scene.And then the peak ratio is taken as confidence to synthesize the depth images. Finally, the fusion depth map is filtered by a combined filter with structural consistency, and a high precision depth map is obtained. The experimental results in the virtual data and real data show that, compared with other advanced algorithms, the depth images obtained by the proposed method are more accurate, less noise, and the edges are more clearer.
作者
张旭东
李成云
汪义志
熊伟
ZHANG Xu-dong;LI Cheng-yunt;WANG Yi-zhi;XIONG Wei(School of Computer and Information,Hefei University of Technology,Hefei 230009)
出处
《控制与决策》
EI
CSCD
北大核心
2018年第12期2122-2130,共9页
Control and Decision
基金
国家自然科学基金项目(61403116)
中国博士后基金项目(2014M560507)
关键词
光场相机
深度估计
全光场图像
遮挡
多线索融合
联合滤波器
light field camera
depth estimation
light field image
occlusion
multi-clue fusion
joint filter
