Omnidirectional Media Format
More actions
Omnidirectional Media Format (OMAF) is an international standard for storing, delivering and presenting 360-degree video and associated media. Published as ISO/IEC 23090-2, Part 2 of MPEG-I, it covers video, still images, audio and timed text, including presentations with several camera viewpoints and overlays.[1] OMAF supplies the format and signaling around media encoded with existing codecs. It is a systems standard rather than a separate video compression codec.[2]
Media structure and rendering
An omnidirectional picture represents directions around a viewing position. To encode that picture as a rectangular image, a production can use equirectangular projection or a cubemap. A cubemap places the surrounding scene on six cube faces. OMAF describes the projection and coordinate system so that a player can interpret the decoded picture. The player generates a viewport corresponding to the current viewing direction, rather than displaying the entire projected rectangle as a flat screen.[3]
Region-wise packing describes how rectangular regions of a projected picture have been arranged in the coded picture. Metadata records region positions, sizes and rotations. A renderer can use that information to place decoded regions back onto the intended spherical or cube surface. Packing and projection therefore describe different stages: one concerns the arrangement of pixels for coding, while the other maps image content to viewing directions.[4]
Timed media use extensions to the ISO base media file format (ISOBMFF). Streaming provisions include metadata for Dynamic Adaptive Streaming over HTTP (DASH) and delivery using MPEG Media Transport.[3] Image extensions use the High Efficiency Image File Format (HEIF). An OMAF media profile combines codec requirements with constraints on signaling and file encapsulation; different profiles cover video, images, audio and timed text.[5]
Viewport-dependent delivery
OMAF supports both delivery independent of viewing direction and delivery adapted to the viewport. In a viewport-dependent system, head tracking can guide which encoded streams or tiles the player requests.[4] Quality-ranking metadata describes the relative quality of image or sphere regions, while initial-viewing-orientation and recommended-viewport metadata serve different purposes: setting a starting direction and describing a preferred view over time.[5]
A tiled HEVC presentation can deliver regions at different resolutions and combine selected tiles into one compressed bitstream before decoding. This allows more detail in the visible region than elsewhere. Fraunhofer HHI's 2019 browser implementation resolved extractor tracks, which identify data in other tracks, and repackaged the result for Media Source Extensions. Its renderer used OMAF packing metadata with WebGL. The demonstration was tested in Safari 12.02; that historical result does not establish support in every browser.[4]
Viewpoints and overlays
A viewpoint corresponds to media from an omnidirectional camera. Multiple viewpoints let a presentation offer different camera locations, such as positions around a sports venue. Metadata can describe their spatial relationships and permissible switches. Authors can determine which viewpoints become available, how switching affects the timeline and which viewing direction follows a transition.[6]
Overlays place additional visual media over a background video, image or viewport. A viewport-relative overlay stays with the viewing screen; a sphere-relative overlay occupies a position in the surrounding presentation. Examples described in the standard's development literature include score displays, supplementary video and interactive controls. Signaling can specify an overlay's position, size, opacity, priority and permitted interactions, including turning it on or off.[6]
In immersive video terminology, 3DoF+ refers to looking around with limited head translation, such as movement while seated.[3] The second edition describes position-dependent rendering of overlays relative to the omnidirectional background. This specific capability and switching between camera viewpoints should be distinguished from reconstructing every view throughout an unrestricted space.[5] MPEG Immersive Video is a separate MPEG-I standard, ISO/IEC 23090-12, concerned with immersive video representation for viewing positions within a limited volume.[7]
Editions
The ISO/IEC publication dates differ from technical completion milestones. The first edition's technical work was completed in October 2017 and the second edition's in October 2020.[5] Their publication history is:[8]
| Edition | Publication date | Scope or additions |
|---|---|---|
| ISO/IEC 23090-2:2019, first edition | 14 January 2019 | Three degrees of freedom (3DoF): changing viewing orientation around a single camera position. Translation alone does not select a different captured view.[8] |
| ISO/IEC 23090-2:2021, second edition | 7 July 2021 | Multiple viewpoints, overlays and additional tiling profiles.[5] |
| ISO/IEC 23090-2:2023, third edition | 9 June 2023 | Adds support for Versatile Video Coding (VVC).[9] |
Implementations and compatibility
Nokia's published Creator/Player implementation demonstrates a subset of the second edition. Its 2.2.0 release, dated 12 October 2021, added multiple viewpoints, overlays and the simple tiling profile. The documentation identifies omissions, including 3D mesh overlays, and states that bitrate adaptation is disabled in the player. These are implementation limits, not limits on every OMAF service.[2]
Compatibility depends on the selected media profiles, rendering features and implementation. OMAF's profile and toolset signaling identifies which features a presentation uses; an implementation need not provide every feature in the standard.[5]
References
- ↑ "ISO/IEC 23090-2:2023: Information technology, Coded representation of immersive media, Part 2: Omnidirectional media format". International Organization for Standardization. 2023-06. https://www.iso.org/standard/84636.html. Retrieved 2026-10-21.
- ↑ 2.0 2.1 "Omnidirectional MediA Format (OMAF): Nokia implementation". Nokia Technologies. https://github.com/nokiatech/omaf. Retrieved 2026-10-21.
- ↑ 3.0 3.1 3.2 "360 degrees and 3DoF+ video". MPEG Workshop on Coding Technologies for Immersive Audio/Visual Experiences. 2019-07-10. https://mpeg.chiariglione.org/sites/default/files/events/4.%20Workshop%20presentation%203DoF%2B%20rev3.pdf. Retrieved 2026-10-21.
- ↑ 4.0 4.1 4.2 "HTML5 MSE Playback of MPEG 360 VR Tiled Streaming". Dimitri Podborski and colleagues, ACM MMSys 2019, author manuscript. 2019-04-23. https://arxiv.org/abs/1903.02971v2. Retrieved 2026-10-21.
- ↑ 5.0 5.1 5.2 5.3 5.4 5.5 "Omnidirectional MediA Format (OMAF): Toolbox for Virtual Reality Services". Sachin Deshpande and Miska M. Hannuksela, IEEE CSCN 2021, accepted manuscript. 2022-03-02. https://arxiv.org/abs/2203.01183. Retrieved 2026-10-21.
- ↑ 6.0 6.1 "Multi-viewpoint and overlays in the MPEG OMAF standard". Igor D.D. Curcio, Kashyap Kammachi Sreedhar and Sujeet S. Mate, ITU Journal: ICT Discoveries. 2020-05-18. https://www.itu.int/dms_pub/itu-s/opb/journal/S-JOURNAL-ICTF.VOL3-2020-1-P03-PDF-E.pdf. Retrieved 2026-10-21.
- ↑ "MPEG-I: Immersive Video". MPEG. https://www.mpeg.org/standards/MPEG-I/12/. Retrieved 2026-10-21.
- ↑ 8.0 8.1 "ISO/IEC 23090-2:2019: Publication details and edition history". International Electrotechnical Commission. https://webstore.iec.ch/en/publication/64564. Retrieved 2026-10-21.
- ↑ "MPEG-I: Omnidirectional Media Format". MPEG. https://www.mpeg.org/standards/MPEG-I/2/. Retrieved 2026-10-21.