LAUSR

Abstract Depth segmentation has attracted much attention in fields ranging from industrial monitoring to image understanding. However, it remains difficult to segment objects positioned at different depths with high speed and high accuracy. This paper proposes a novel depth segmentation scheme based on the phase-shift invariance of the segmentation lines and a simple optimization framework that exploits simple image processing algorithms to enhance its performance in noisy environments. Without the need for precalibration or complementary cues, this approach performs depth segmentation effectively by changing the pattern sequence selectively during the step of postprocessing. First, the structural configuration and special properties of the phase-shifting algorithm are described, following which an optimization framework is established to allow data matting and labeling. Simulation and experimental results show that this approach can perform depth segmentation efficiently despite the influence of environmental noise. Moreover, the processing is of low cost, only needing to detect the intersections of discontinuities among three different wrapped phase maps. Most significantly, this method is robust to variations in environmental noise, camera exposure, and object color and texture.