Light field compression using depth image based view synthesis

This paper describes a novel light field compression scheme using a depth image-based view synthesis technique. A small subset of views is compressed with HEVC inter coding and then used to reconstruct the entire light field. The residual of the whole light field can be then restructured as a video sequence and encoded by HEVC inter coding. Experiments show that our scheme significantly outperforms a similar view synthesis method which utilizes convolutional neural networks, and does not require training with a large dataset of light fields as required by deep learning techniques. It also outperforms as well the direct encoding of all the light field views.