Low bit-rate compression of video and light-field data using coded snapshots and learned dictionaries.