基于光场分析的多线索融合深度估计方法

Depth Estimation from Light Field Analysis Based Multiple Cues Fusion

受人类视觉深度感知机理的启发,结合最近光场分析理论的进展,文中提出了一种融合多种深度线索的全局一致深度估计方法.该方法首先利用摄像机阵列获取光场数据,然后通过合成孔径成像方法获得指定深度的重投影光场,并从中提取表征深度变化不同维度的模糊与视差线索.接着文中采用光场分析方法对比了模糊与视差线索的适用情况,设计了多深度线索的融合算法,以实现不同深度线索的优势互补.进而为了获得精确一致的深度结果,文中在马尔可夫随机场模型的基础上,提出了一种自适应平滑约束的全局能量函数.最终,文中利用图割算法最小化全局能量函数,获得了平滑的高精度深度估计结果.文中分别在虚拟数据和真实数据上测试了所提出的方法,与单一深度线索和局部深度估计方法相比,文中方法能结合不同深度线索的优势并获得更加鲁棒的深度结果.

Inspired by the depth perception mechanism of the human visual system,this paper proposes a novel globally consistent depth estimation method from defocus blur and stereo disparity depth-cues.According to the recent progress of light field theory,we can simulate the focus and defocus depth cue by using synthetic aperture photograph technique.Firstly,the defocus blur and stereo disparity depth cues are extracted from light field data sets,which is acquired by using a camera array system.Then based on the characteristic of the two depth cues,we design a fusion algorithm for the light field depth estimation.To acquire the globally consistent structural depth result,we introduce an adaptive weighted smoothing function in Markov random field framework.Finally,the global energy is minimized by graph cut algorithm,which leads a consistent and precise depth estimation.We test the proposed method on both virtual data and real data,the experimental results have shown that our method can take advantage of the two depth-cues and obtain more robust depth estimation.