Distance estimation based on light field geometric modeling

Abstract

In this paper, geometric optical models are proposed that can accurately estimate the distances of object planes in a light field image. Treating an arbitrary point on an object plane as an off-axis point light source, the single-point-light-source model, which consists of two relaying models, is established to describe the propagation of light rays from the off-axis point light sources to the image sensor. Based on the relaying models, the distances of object planes can be estimated by deriving the relationship between the positions of light rays impinging on the image sensor and incident angles of light rays entering into the main lens. Furthermore, taking into account the realistic imaging, the so-called line-source model is established by extending the off-axis point light sources to general line sources. The performances of the proposed models are compared with existing distance estimation methods under two optical configurations of plenoptic 1.0 cameras and results demonstrate that the proposed methods improve estimation accuracy by an average of 2.315%/4.675% for point light sources and 2.77% for a line source.