Edge Segmentation in Images of a Focused Plenotic Camera

Abstract

In this article we present a method for 3D segmentation which is optimized for the recordings of a focused plenoptic camera. Since, caused by the imaging concept of a plenoptic camera, depth information is only received for regions of high contrast our approach is based on retrieving 3D edge segments out of the RGB-D data acquired from a plenoptic camera. In our method the depth data is filtered by a nonlinear filter. Afterwards, a 2D edge segmentation is performed to the RGB image. The 2D segments are projected to the 3D space by applying a weighted least squares estimate to the filtered depth data.