A lightweight foveation codec for VR

Summary

We present a foveated VR video codec, consisting of an image pre-reduction (warp) that can be combined with conventional video codecs such as H.265. Initial live testing between two machines several kilometers apart gave no visible artifacts at a compression ratio of 5:1. The warp’s visibility impact was subsequently evaluated on various real-world test scenes, yielding a performance envelope with respect to acceptable latencies and compression ratios. The compression ratio was approximately confirmed, with a trade-off zone ranging from 4:1 to 7:1, depending on system latency. Ultimately, we argue that this approach can reasonably be used at least at a compression ratio of 5:1, leading to additional total bandwidth savings of 34 % over an assumed static warp with a compression ratio of 3:1 (up to 48 % at a compression ratio of 7:1 that worked for some users); is practicably implementable on current architectures and can improve visual quality over static foveation.