Matroska/WebM in MPEG DASH

This document defines how to use Matroska/WebM with the MPEG DASH adaptive streaming system defined in ISO-IEC_23009-1 [DASH].

Matroska and WebM share the same file structure, Matroska having some extra features. WebM is also restricted to the VP8 video codec and Vorbis audio codec. The Matroska specification can be found at http://www.matroska.org/technical/specs/index.html [MATROSKA] and the WebM specification can be found at http://www.webmproject.org/code/specs/container/ [WEBM]. There is also an overview of all the structures found in Matroska/WebM for beginners at http://www.matroska.org/technical/diagram/index.html. The MPEG DASH terminology appears in bold and the Matroska terminology appears in italic. For example Segment in DASH and Segment in Matroska.

1 Segment formats for Matroska/WebM

1.1 Preliminaries: Refinements of generic concepts

1.1.1 Subsegments

Media subsegments are defined as one or more consecutive Clusters.

1.1.2 Media stream access points

Media stream access points correspond to a CuePoint contained in a Cues. Typically CuePoints reference Blocks that are marked as key frames within a single stream.

For WebM all SAPs shall be of type 1. In this case a SAP shall not reference any Blocks preceding the SAP and the SAP shall be the first frame rendered.

1.1.3 Segment Index

The Segment Index corresponds to the Cues.

1.2 Initialization Segment Format

The Initialization Segment shall contain the EBML header, Segment header, Segment Information and Tracks. The Initialization Segment may contain other level1 elements and padding.

If the Media Segment contains a Cues that is placed after the Clusters than the Initialization Segment shall contain a SeekHead with a reference to the Cues.

The Initialization Segment shall not contain Clusters or Cues.

1.3 Media Segment Types

Media Segments can be of three different types: Basic Media Segments, Self-Initializing Media Segments, and Self-Initializing Indexed Media Segments.

1.3.1 General format type

- All Media Segments must contain one or more Clusters.

1.3.2 Basic Media Segment

- The Media Segment shall contain only one or more Clusters.

- The Initialization Segment that references these Media Segments should have a Segment with an unknown size (-1).

1.3.3 Self-Initializing Media Segment

- Each Media Segment shall contain an Initialization Segment.

- The Initialization Segment shall be placed before the Clusters.

1.3.4 Self-Initializing Indexed Media Segment

- Each Media Segment shall contain an Initialization Segment.

- The Initialization Segment shall be placed before the Clusters.

- Each Media Segment shall contain a Cues that references the Clusters in the Media Segment.

- The Cues shall not be part of the Initialization Segment.

- The Cues shall be placed after the Initialization Segment.

- If the Cues is placed after the Clusters then a SeekHead shall be contained in the Initialization Segment with a reference to the Cues.

1.4 Media Presentation based on Matroska/WebM formats

1.4.1 MIME types

The MIME types for Segments are as follows:

• video/webm for WebM video files,

• audio/webm for WebM audio-only files

• video/x-matroska for Matroska video files

• audio/x-matroska for Matroska audio-only files

• video/x-matroska-3d for Matroska files with a stereoscopic video trackRFC 4281

The MIME type for codecs are as defined in RFC 4281 [MIMETYPE]. For WebM only video/vp8 and audio/vorbis are possible.

1.4.2 General

- In the case a Representation contains only one Media Segment then the Media Segment must be a Self-Initializing Indexed Media Segment.

- An Index Segment shall not be present. However an @indexRange attribute may be present to signal the byte range for a Segment Index within a Media Segment.

- If an Initialization Segment is contained within a Self-Initializing Indexed Media Segment then the @range attribute of the Initialization Segment shall be present to signal the byte range for the Initialization Segment within the Media Segment.

1.4.3 Authoring Rules for specific MPD flags

1.4.3.1 Segments starting with media stream access points

- If the @startWithRAP attribute is not set to '0', the conditions in 5.5.3.2 [DASH] shall apply.

1.4.3.2 Subsegments starting with media stream access points

- If the @subsegmentStartsWithRAP attribute is not set to '0', the conditions in 5.5.3.2 [DASH] shall apply.

- Each subsegment shall start on a Cluster boundary.

- The first Block in the subsegment shall be a key Block.

1.4.3.3 Subsegment Alignment

- If the @segmentAlignmentFlag is set to 'true', the conditions in 5.5.3.2 [DASH] shall apply.

- The Cues of each Representation within an AdaptationSet shall have the same timecodes.

- Each CuePoint within a Cues should reference the first Block of the TrackNumber within a Cluster.

1.4.3.4 Bitstream switching

- If the @bitstreamSwitchingFlag is set to 'true', for a set of Representations within an AdaptationSet, the conditions in 5.5.3.2 [DASH] shall apply and the Bitstream Switching Segment shall not be present.

- For a set of Representations within an AdaptationSet, the TrackNumber, CodecID and CodecPrivate shall contain the same value.

- The codec shall support transparent resolution switching. (e.g. VP8)

2 Matroska/WebM On-Demand profile

2.1 General

- The conditions in 8.3.1 [DASH] shall apply.

- For WebM the On-Demand profile is identified by the URN “urn:mpeg:dash:profile:webm-on-demand:2012”.

2.2 Media Presentation Description Constraints

Most of the constraints are the same constraints as those outlined in 8.3.2 [DASH].

- The rules for the MPD and the segments as defined in Media Presentation based on Matroska/WebM formats shall apply.

- Representations not inferred to have @profiles equal to the profile identifier as defined in On-Demand General may be ignored.

- MPD@type shall be “static”.

- The Subset element may be ignored.

- neither the Period.SegmentList element nor the Period.SegmentTemplate element shall be present

- if either the AdaptationSet.SegmentList or the AdaptationSet.SegmentTemplate element is present in an AdaptationSet element then this AdaptationSet element may be ignored.

- if either the Representation.SegmentList or the Representation.SegmentTemplate element is present in a Representation element then this Representation element may be ignored.

- if the Representation element does not contain a BaseURL element then this Representation element may be ignored.

- AdaptationSet elements with AdaptationSet@subsegmentAlignment not present, or set to 'false' may be ignored.

- Representation elements with @subsegmentStartWithSAP value (either supplied directly or inherited from the containing AdaptationSet) that does not equal 1 may be ignored if the containing AdaptationSet contains more than one Representation.

- Elements using the @xlink attribute may be ignored from the MPD. The Representations conforming to this profile are those not accessed through an AdaptationSet that uses an @xlink.

2.3 Segment Format Constraints

- Each Representation shall contain one Segment that complies with the Self-Initializing Index Media Segment.

3 Matroska/WebM Live profile

TODO(fgalligan)

4 References

[DASH] Dash specification

www.itscj.ipsj.or.jp/sc29/open/29view/29n12313t.doc

[MATROSKA] Matroska specification

            http://www.matroska.org/technical/specs/index.htm

[MIMETYPE] RFC mimetypes for codecs

http://www.ietf.org/rfc/rfc4281.txt

[WEBM] WebM specification

http://www.webmproject.org/code/specs/container/