AVIF Brand and AV1 Intra-Coding Compliance

This article explains how the avif brand identifies image items encoded using AV1 intra-coding within the ISO Base Media File Format (ISOBMFF). It breaks down the role of the File Type Box (ftyp), codec configuration parameters, and the structural requirements that guarantee an image relies exclusively on self-contained AV1 intra frames without temporal prediction.

The Role of the File Type Box (ftyp)

In the ISOBMFF structure inherited by AVIF from the High Efficiency Image File Format (HEIF), file conformity begins at the File Type Box (ftyp). This box contains two critical fields:

• major_brand: Signals the primary specification the file conforms to and identifies the best parsing engine.
• compatible_brands: A list of Four-Character Codes (FourCCs) indicating other specifications the file adheres to.

Setting the major_brand or a value in compatible_brands to 'avif' (in ASCII: 0x61 0x76 0x69 0x66) formally declares that the primary resource is an AV1-encoded still picture.

Signatures of AV1 Intra-Coding

The 'avif' brand explicitly requires the primary image item to consist solely of AV1 intra-coded data. Unlike AV1 video streams, an AV1 still picture inside an AVIF file cannot rely on temporal inter-frame prediction. Compliance is designated by several specific constraints:

1. Frame Type Enforcement: The AV1 bitstream mapped to the primary item must consist of an AV1 Still Picture, which is an Open Bitstream Unit (OBU) sequence containing an OBU_SEQUENCE_HEADER followed by a single frame. In this frame's uncompressed header, the frame_type must be set to KEY_FRAME or INTRA_ONLY_FRAME. Inter-coded frames (INTER_FRAME or SWITCH_FRAME) are invalid under the 'avif' brand.

2. The AV1 Configuration Item Property (av1C): Under HEIF rules, items in an AVIF container are described by item properties in the iprp (Item Properties) box. The image item associated with the 'avif' brand must be linked to an av1C property box. This property includes:

   • seq_profile: Indicates the AV1 profile (e.g., 0 for Main, 1 for High, 2 for Professional).
   • seq_level_idx_0: Indicates the level limit.
   • Color configuration: Specifies bit depth (8, 10, or 12 bits), subsampling format (e.g., 4:2:0, 4:4:4), and color primaries.

3. Absence of Temporal References: Because the 'avif' brand represents a single still image item rather than an image sequence (which uses the 'avis' brand), the decoder is not required to initialize or maintain reference frame buffers for future frames. The image is parsed and reconstructed solely from intra-prediction tools, including directional intra modes, recursive filtering, and cross-component prediction.

Distinction from the 'avis' Brand

The AVIF specification differentiates single items from sequences:

• 'avif': Designates static images, including derived images such as image grids and alpha auxiliary items, where all components are intra-coded AV1 items.
• 'avis': Designates image sequences (such as animations or bursts), which are stored as tracks rather than single items and may allow inter-frame prediction across temporal units.

When a parser encounters 'avif' in the ftyp box, it treats the primary payload as a standalone, intra-coded frame compliant with the AV1 Still Picture standard, enabling immediate decoding without temporal dependencies.