Initial depth estimation using EPIs and structure tensor

Abstract

Light field cameras are enlarged in the field of computational photography because of the increased information they capture compared to traditional cameras. Moreover, the estimated depth is widely used for photographic editing. While, the intrinsic information of the light field has the advantage to allow disparity computation without the image calibration and rectification steps required in classic binocular stereo or multi-view algorithms. In this paper, we propose an initial depth estimation method using epipolar plane images (EPIs) and structure tensor, which appropriately applicable to the sub-aperture light field images. First, the sub-aperture images are displaced using raw light field image data. We analyze the sub-aperture images to EPIs and compute depth and corresponding information using the structure tensor on EPIs. Finally, the experiment results show that an initial disparity map of light field image is reconstructed.