AVIF Parsing: Still Image vs Image Collection

An AVIF parser distinguishes between a single still image and an image collection by analyzing the structural metadata defined in the ISO Base Media File Format (ISOBMFF) and the High Efficiency Image File Format (HEIF) specifications. By inspecting key boxes within the file—specifically the File Type Box (ftyp), the Meta Box (meta), the Primary Item Box (pitm), and the Item Information Box (iinf)—the parser determines whether the file contains one primary visual item with optional auxiliary data or multiple independent images meant to be treated as a collection.

1. Evaluating Brands in the File Type Box (ftyp)

The parser begins at the ftyp box to identify the file's major and compatible brands:

• avif: Indicates support for static AVIF items.
• avis: Indicates an AVIF image sequence (typically animated content stored across video-style tracks).

While ftyp signals general format support, it does not fully resolve whether a file contains a single still or an image collection, as both can share the avif structural brand. To make this determination, the parser must inspect the structural layout within the meta box.

2. Inspecting the Primary Item Box (pitm)

Inside the meta box, the parser checks for the presence of a Primary Item Box (pitm).

• Single Still Image: The pitm box specifies a valid item_ID representing the primary display item (an AV1 image item of type av01 or a derived image such as a grid).
• Image Collection: A collection may either omit the pitm box entirely (signaling no default master image) or include a pitm box merely to designate a cover/representative image for the broader set. The presence of pitm alone is not sufficient to confirm a single image; the parser must evaluate how other items relate to it.

3. Analyzing the Item Information Box (iinf) and Item Types

The parser inspects the iinf box, which catalogues all items in the file via Item Info Entry (infe) boxes. The parser filters items by their item_type:

• Master Image Items: Items with the type av01 (AV1 encoded frames) or derived types like grid or iden.
• Auxiliary Items: Non-displayable items such as alpha transparency channels, depth maps, or gain maps (often denoted by auxl or associated metadata).
• Hidden Items: Items marked with the item_protection_index or hidden flags.

If the parser finds only one top-level, non-auxiliary image item, the file is classified as a single still image. If multiple independent master image items exist, the parser looks further to identify their relationships.

4. Resolving Item References (iref)

To confirm whether multiple items represent a single image or an image collection, the parser checks the Item Reference Box (iref):

• Single Still Image Dependencies: Multiple items often compose a single image. For instance, an alpha channel references the master image with an auxl reference type, or a thumbnail references it with a thmb reference type. Derived images (such as a tiled grid) will reference multiple sub-image tiles through a dimg (derived image) reference. Because these items depend on or support a single output canvas, the parser treats the entire structure as a single still image.
• Image Collection: If multiple independent master image items exist that are not consumed by a derived image (like a grid) and are not auxiliary references (like alpha or thumbnails), the parser identifies the file as an image collection. In this scenario, each master item represents an autonomous image.

5. Meta Items vs. Movie Tracks (moov)

The parser checks if the file uses track-based storage via a Movie Box (moov):

• Item-based (Meta Box): AVIF still images and static image collections reside exclusively in the meta box as individual items.
• Track-based (Movie Box): If visual data is packaged in tracks (trak), the file is categorized as an image sequence (animation) rather than a static collection.