Method for Efficient CPU-GPU Streaming for Walkthrough of Full Motion Lightfield Video

Abstract

Lightfield video, as a high-dimensional function, is very demanding in terms of storage. As such, lightfield video data, even in a compressed form, do not typically fit in GPU or main memory unless the capture area, resolution or duration is sufficiently small. Additionally, latency minimization--critical for viewer comfort in use-cases such as virtual reality--places further constraints in many compression schemes. In this paper, we propose a scalable method for streaming lightfield video, parameterized on viewer location and time, that efficiently handles RAM-to-GPU memory transfers of lightfield video in a compressed form, utilizing the GPU architecture for reduction of latency. We demonstrate the effectiveness of our method in a variety of compressed animated lightfield datasets.