基于随机遮挡孔径的光场图像混叠检测方法

Aliasing Detecting of Light Field Images Based on Random Masked Aperture

现有光场采集系统存在角度信息欠采样的问题,由此引起的图像混叠效应降低了光场图像质量.本文在2D光场框架下分析了光场图像混叠的空域模型,并且提出一种变换离散孔径采样密度的混叠检测方法.该方法通过计算随机遮挡孔径(Random masked aperture,RMA)成像点集的变异系数进行混叠检测,特点在于无需已知场景深度和纹理的先验信息.以平面相机阵列为采集平台,本文在多组真实光场数据集上进行了方法的验证,并且在检测结果基础上对混叠效应进行了修正.

In a light field imaging system, the angular down-sampling problem will lead to severe aliasing effects, which can significantly deteriorate the quality of a light field image. To address this problem, first we model the causes of aliasing effects in a 2D light field framework. Then, we propose a random masked aperture （RMA） based aliasing detecting algorithm. We use the coefficient of variations of imaging set derived from random masked aperture as an aliasing metric. Most importantly, the proposed algorithm is free of depth estimation and texture independent. We have validated the proposed algorithm on several groups of real light field datasets, which are acquired by using a planar camera array system. Finally, we alleviate the aliasing artifacts by employing the detecting results.