Efficient Depth Estimation and Fast 3D Reconstruction Using Light Field Imagery

Abstract

Three-dimensional reconstruction is one of the classical problems in computer vision, and its application area is widely used , which has been a hot spot for research in related fields. And the accuracy and speed of 3D reconstruction depends on the estimation of scene depth information. With the development of light field imaging technology, it is more and more convenient to acquire light field images, which contain four-dimensional information and are beneficial to the accurate estimation of scene depth information. The application of deep learning in light field image depth estimation improves the speed and accuracy of light field image depth estimation, and further enables the 3D reconstruction of the scene. In this paper, we study the use of light-field images combined with deep learning for scene depth estimation, and finally realize the near- field fast 3D reconstruction.