A novel algorithm was developed to estimate scene depth by uniquely exploiting plenoptic image data. Plenoptic images allow multiple perspectives to be extracted from a single exposure from a single lens. Depth from disparity was calculated for every available perspective view using normalized cross correlation. This resulted in a large number of unique disparity maps, which were then combined in a least-squares sense, weighted by confidence coefficients to account for measurement uncertainty. The averaged disparity map was then transformed into a depth map using geometric optics. The accuracy of the algorithm was tested with a plenoptic camera on a small static target imaged throughout the depth of field. Depth was recovered with a standard deviation less than 5 mm, validating the algorithm.
Correlation-Based Depth Estimation with a Plenoptic Camera
AIAA journal ; 55 , 2
2017
Aufsatz (Zeitschrift)
Englisch
Correlation-Based Depth Estimation with a Plenoptic Camera
Online Contents | 2016
|Correlation-based Depth Estimation with a Plenoptic Camera (AIAA 2016-0136)
British Library Conference Proceedings | 2016
|AIAA | 2017
|