Understanding MV-HEVC: Advances in 3D Video Encoding and V265 Support for Apple Vision Pro

Video has become ubiquitous, and users now demand immersive 3D experiences beyond traditional 2D content. 3D video simulates binocular vision to provide depth perception, and several imaging technologies such as holographic projection, glasses‑free 3D screens, stereoscopic cinema, and VR head‑sets are used to deliver it.

Most current 3D video is encoded using generic codecs in a side‑by‑side (SBS) format, which merges left and right eye views into a single frame, leading to redundant information and limited compression efficiency.

HEVC lacks inter‑view prediction, so left‑right view data cannot be jointly predicted, resulting in missed opportunities for bitrate reduction.

MV‑HEVC (MultiView‑HEVC) extends HEVC by introducing a LayerId in the NALU header to differentiate view layers, allowing inter‑layer prediction and new reference structures that eliminate inter‑view redundancy.

The standard also expands VPS, SPS, and PPS to support separate parameter sets for each layer, reducing duplicated metadata.

Inter‑Layer prediction introduces challenges such as potential division‑by‑zero when motion‑vector scaling factors are zero; MV‑HEVC resolves this by marking all inter‑layer references as long‑term reference frames.

Encoding tests show that, in 3D movie scenarios, a V265 encoder with MV‑HEVC extensions saves over 50% bitrate on the auxiliary view and more than 20% overall compared to conventional HEVC, enabling up to 20% bandwidth reduction at equal visual quality.

Apple Vision Pro’s hardware already supports MV‑HEVC decoding, and the V265 implementation can be integrated into FFmpeg for broader adoption.

In summary, adding MV‑HEVC syntax and modes gives Tencent’s V265 encoder 3D video capabilities, delivering notable compression gains while leveraging the ecosystem’s hardware decoding support to foster the growth of 3D video applications.