AV1 Video

AV1 is a video coding format defined by the Alliance for Open Media with profiles, levels, and optional film-grain signaling.

AV1 is the video coding format defined by the Alliance for Open Media. Its bitstream contains instructions for reconstructing picture samples, maintaining reference state, and preparing visible output.

One useful AV1 distinction is that the picture retained for later prediction and the picture prepared for display can pass through different operations.

OBUs, frame headers, and superblock partitions

AV1 organizes coded data into Open Bitstream Units (OBUs). Sequence information establishes the applicable profile, sample format, and other stream parameters. Frame information describes the picture's reconstruction choices, reference updates, and output behavior. Tile data supplies the coded regional information.

A decoder needs the relevant headers as well as the block payloads. The picture isn't described by an isolated list of pixel values.

Its block organization can start with 64 × 64 or 128 × 128 superblocks. Regions are partitioned further to fit the picture structure. The chosen partitions define where prediction and correction decisions apply.

Construct the scene and its correction

Intra modes use reconstructed information within the picture. Inter modes use retained decoded pictures with motion information, including permitted compound predictions from multiple views.

A compound predictor can combine corresponding estimates of a region. Masks or weights can give different contributions where the coding mode permits them. The aim is to leave a correction that can be described efficiently.

Transform and quantization operations represent that remaining correction. The decoder recovers coefficients, constructs the correction, and adds it to the predictor. Quantization settings control the approximation in ordinary lossy AV1, while the partition and prediction choices control how effectively the input is represented.

Filter the reconstructed picture inside the coding loop

AV1 includes several reconstruction-filtering stages. Deblocking addresses selected block-boundary behavior. CDEF, the Constrained Directional Enhancement Filter, identifies a local direction and constrains its filtering so the operation can address artifacts while respecting that structure. Loop restoration supplies additional signaled processing of reconstructed regions.

The result used for reference updates is the applicable loop-restored picture. Later prediction needs these samples, so the encoder must locally reconstruct corresponding reference state.

This separates coding-loop work from optional presentation processing. A filter used outside the reference loop can change what is shown without changing the samples later blocks predict from.

Add grain to output when it is signaled

Some AV1 streams describe film grain with parameters rather than coding every original grain sample as picture texture. An encoder can represent a cleaner scene and supply information for synthesizing grain in output processing.

Those parameters describe such things as how grain strength varies with brightness, how nearby grain samples relate, and the generation seed. The decoder uses the specified process to produce the intended grain-like appearance.

Grain is applied to output samples; the loop-restored reference picture remains the predictor state. Keeping the newly synthesized pattern outside that reference avoids making future scene prediction depend on that particular display grain realization.

The film-grain explanation follows the preservation and modeling tradeoff. The metadata describes an appearance model, rather than an exact archive of every grain sample captured in the source.

Decoded pictures and visible pictures

AV1 can decode a hidden reference picture and later show a retained picture through its output instructions. A coded unit count therefore need not equal the number of new visible camera moments.

The container's timeline and the format's picture/output instructions cooperate to schedule the visible result. In an Emby conversion, the chosen decoded output needs to include the applicable grain and presentation interpretation for the result being encoded or displayed.

Reference details for the format

Detail Meaning
Main profile, sequence profile 0 8-bit or 10-bit, 4:2:0 or monochrome sample permissions
Other profiles Additional permitted sample formats
Level and tier Workload and coded-rate constraints
libaom, SVT-AV1, hardware encoder Implementations choosing an AV1 representation
Grain parameters An optional output-synthesis description

An encoder's search and rate-control decisions determine which valid representation is produced. A decoder follows its AV1 syntax and reconstruction rules; it does not need to be the implementation that created the stream.

See Also

References