Depth-independent segmentation of macroscopic three-dimensional objects encoded in single perspectives of digital holograms

We present a technique for performing segmentation of macroscopic three-dimensional objects recorded using in-line digital holography. We numerically reconstruct a single perspective of each object at a range of depths. At each point in the digital wavefront we calculate variance about a neighborhood. The maximum variance at each point over all depths is thresholded to classify it as an object pixel or a background pixel. Segmentation results for objects of low and high contrast are presented.