Range Finding with a Plenoptic Camera

摘要

Abstract : The plenoptic camera enables simultaneous collection of imagery and depth information by sampling the 4D light field. The light field is distinguished from data sets collected by stereoscopic systems because it contains images obtained by an N by N grid of apertures, rather than just the two apertures of the stereoscopic system. By adjusting parameters of the camera construction, it is possible to alter the number of these subaperture images,' often at the cost of spatial resolution within each. This research examines a variety of methods of estimating depth by determining correspondences between subaperture images. A major finding is that the additional 'apertures' provided by the plenoptic camera do not greatly improve the accuracy of depth estimation. Thus, the best overall performance will be achieved by a design which maximizes spatial resolution at the cost of angular samples. For this reason, it is not surprising that the performance of the plenoptic camera should be comparable to that of a stereoscopic system of similar scale and specifications. As with stereoscopic systems, the plenoptic camera has its most immediate, realistic applications in the domains of robotic navigation and 3D video collection.