Abstract

This paper presents an innovative technique for 3D depth estimation from a single holoscopic 3D image (H3D). The image is captured with a single aperture holoscopic 3D camera, which mimics a fly’s eye technique to acquire an optical 3D model of a true 3D scene. The proposed method works by extracting of optimum viewpoints images from a H3D image and it uses the shift and integration function in up-sampling the extracted viewpoints and then it performs the match block functions to match correspondence features between the stereoscopic 3D images. Finally, the 3D depth is estimated through a smoothing and optimizing process to produce a final 3D depth map. The proposed method estimates the full 3D depth information from a single H3D image, which makes it reliable and suitable for trend autonomous robotic applications.

© 2017 Optical Society of America