Acquisition of light field images & videos: Capturing light rays

Abstract:

Even though there are many different ways acquiring rays from a light field (e.g., [5]), we will in this chapter restrict ourselves to three major architectural implementations: Plenoptic cameras: Due to the long history of photography and the availability of devices, the first implementations of plenoptic imagery have been by adding so-called MLAs (microlens arrays) to standard 2D-cameras. Such hand-held plenoptic photography [3] reused 2D sensors to capture a set of SAIs (sub-aperture images), so that the resolution of the sensor is split between the spatial domain (aka the resolution of each individual SAI) and the angular domain (aka the number of such SAIs). - Light field gantries: Based on the insight that light fields for sufficient quality do require a large number of rays (actually the spatial resolution should be comparable to 2D-photography so that the overall data-rate is multiplied by the number of SAIs), light field gantries have been proposed. They use high-resolution 2D-cameras and very fine granular positioning systems (down to the 100 μm-domain) to capture a very large number of SAIs with sufficient resolution each. - Light field arrays: To overcome the limitation of gantries to still image photography (the scene and illumination need to remain constant during the whole capturing process), most of the recent light field captures have been taken by assemblies of cameras in camera-arrays. Different geometries (from planar to spherical arrays) are used. Since arrays offer the most intuitive capture of light fields (the resolution of each SAI is given by the individual camera, and the angular resolution is given by number and spacing of such individual cameras), in this book, we adapt our nomenclature to light field arrays.