MPEG Immersive Video

MPEG Immersive Video (MIV) is a standard for representing and compressing volumetric video so that a viewer can change both viewing position and orientation within a limited region. It is Part 12 of the MPEG-I family, formally ISO/IEC 23090-12. The second edition, ISO/IEC 23090-12:2025, was published in September 2025 and replaced the first edition.

Reviewed 21 October 2026. Official standards records and white papers, original coding paper and verification report About review dates.

Viewing space and XR applications

MIV's six degrees of freedom include rotation and translation. A scene is captured by several real cameras or rendered from several virtual cameras. The representation supplies information for generating views between those camera positions. Its intended viewing space depends on the capture arrangement; the standard describes movement within a constrained range.

This supports motion parallax, in which nearby and distant objects change their relative image positions as the viewpoint moves. Ordinary 360-degree video lets a viewer look around from a fixed position. MIV also represents scene depth for positional viewing. Intended applications include immersive sports viewing, training and telepresence. Playback can use a head-mounted display or a conventional screen whose viewpoint responds to user input.

Representation and coding

Source views

The usual multiview video plus depth representation combines images of a scene with corresponding depth maps. Geometry may come from depth sensors, computer vision or a computer-generated scene. Camera metadata describes the projection and camera pose, allowing samples from different views to be related in three-dimensional space. Perspective, equirectangular and orthographic projections are supported. The capture system and the technique for estimating depth are outside MIV's scope.

Pruning, patches and atlases

Nearby cameras often record the same surfaces. MIV encoders can reduce this duplication before conventional video compression. In the Test Model for Immersive Video (TMIV), selected basic views retain their image data. Additional views are pruned to retain information needed beyond those basic views, including surfaces that were occluded from another camera. The retained regions become rectangular patches, which are packed into larger two-dimensional images called atlases.

The atlas video is compressed by an underlying two-dimensional codec. MIV metadata records each patch's source view and its position and orientation in the atlas. A decoder uses this information to undo the packing and supply geometry and image data for view synthesis. MIV therefore organizes scene information around existing video coding technology; it does not replace all the functions of codecs such as HEVC or VVC.

Decoding and rendering

MIV extends the Visual Volumetric Video-based Coding (V3C) bitstream structure in ISO/IEC 23090-5. It shares that framework with video-based point cloud compression (V-PCC), while specifying different scene representation and processing. Its specification defines the bitstream syntax and required decoding processes. Encoding choices and the final renderer have implementation freedom. A hypothetical reference renderer illustrates reconstruction, but conforming implementations do not have to reproduce that renderer's operations.

Editions and profiles

The first published edition was ISO/IEC 23090-12:2023, issued in August 2023. The second edition reached Final Draft International Standard status at MPEG's January 2025 meeting, before its September publication.

Profiles restrict the tools a bitstream uses and consequently the capabilities needed to decode it. The second edition retains the earlier profiles for compatibility and adds MIV 2 and MIV Simple MPI.

Profile	Main purpose
MIV Main	Multiview video with depth.
MIV Extended	Additional representation options, including transparency and multi-plane imagery (MPI).
MIV Geometry Absent	Omits transmitted geometry, supporting workflows that estimate depth elsewhere.
MIV 2	Supports the additional tools introduced with the second edition.
MIV Simple MPI	Transmits full-plane multi-plane imagery using a single video sub-bitstream.

Second-edition additions include capture-device information, patch margins, background views and static background atlases, colorized geometry, and further depth-quantization tools. MPEG's announcement also describes the Simple MPI profile as allowing playback with a single two-dimensional video decoder.

Research and implementation limits

A 2023 MPEG verification test compared MIV Main and Geometry Absent configurations with an MV-HEVC reference configuration. MIV Main was tested on four scene sequences at target rates of 5, 17, 28 and 40 Mbit/s; Geometry Absent was evaluated on the CBABasketball sequence. Thirty viewers initially participated at RWTH Aachen University, with inconsistent ratings screened out during analysis. Participants rated prerecorded viewpoint paths on two-dimensional displays using an 11-grade quality scale.

The report found overall quality advantages over that reference configuration, with results varying by sequence and rate. It evaluated the complete encoding and rendering configurations using predefined paths. It did not test freely navigated headset viewing or establish a headset-comfort benefit. The report also identified limitations in some source material and saturation where additional bitrate gave little improvement. MPEG's second-edition white paper states that no new verification test was conducted for that edition because its principal changes added functionality.

Capture density, camera placement and depth quality constrain the useful viewing region. Thin structures and hollow shapes can be difficult to capture, and geometry or compression errors can affect synthesized views. Six degrees of freedom describes the supported movement axes, rather than unrestricted travel through a fully recorded world. The TMIV encoder, decoder and renderer support research and development; their algorithms are reference implementations rather than a requirement that every product use the same processing.

Related media formats

Omnidirectional Media Format (OMAF), ISO/IEC 23090-2, is a separate MPEG-I part concerned with omnidirectional media presentation. Its first edition addressed a single camera position, while later work added support for more interactive viewing. Apple Immersive Video is Apple's separately named media format and production workflow for stereoscopic, large-field-of-view playback on visionOS. The shared words "immersive video" do not make it the MPEG MIV standard.