Efficient Depth Estimation for Light Field Images
Wisarut Chantara, Ji-Hun Mun, Yo‐Sung Ho · 2018
Light field images consist of a set of sub-aperture images with slight shifts in their positions. They can be used to obtain the depth map without camera calibration and image rectification. However, since the baseline between sub-aperture images is very narrow, it limits the depth estimation range and accuracy. In this paper, we propose a depth estimation method to solve such problems. After we analyze the sub-aperture images in the form of epipolar plane images (EPIs), we modify the edge confidence measure to test which part of EPIs is good for depth estimation. Then, we apply the modified structure tensor (MST) to compute depth and corresponding information. Finally, we reconstruct a depth map from light field images.