Fast and Accurate Reconstruction of Compressed Color Light Field

Light field photography has been studied thoroughly in recent years. One of its drawbacks is the need for multilens in the imaging. To compensate that, compressed light field photography has been proposed to tackle the tradeoffs between the spatial and angular resolutions. It obtains using only one lens, a compressed version of the regular multi-lens system. The acquisition system consists of a dedicated hardware followed by a decompression algorithm, which usually suffers from high computational time. In our work, we suggest to compress the color channels as well and propose a computationally efficient neural network that achieves high-quality color light field reconstruction from a single coded image. Our approach outperforms existing solutions in terms of recovery quality and computational complexity. We also present a neural network for depth map extraction from the decompressed light field, which is trained in an unsupervised way without the ground truth depth map.