AV1 Maximum Superblock Size Explained
This article explores the superblock architecture within the AV1 video codec, specifically focusing on its maximum superblock size and its role in block partitioning. By examining how AV1 handles macroblock processing compared to older standards, this guide details the structural specifications, partitioning flexibility, and performance advantages of AV1's 128x128 superblocks in modern video compression.
The AV1 (AOMedia Video 1) video codec supports a maximum superblock size of 128x128 pixels. While it also allows an encoder to choose a 64x64 superblock size—the standard maximum used in VP9 and HEVC (H.265)—the 128x128 configuration represents the upper limit designed to optimize high-resolution encoding.
The Role of 128x128 Superblocks
Superblocks serve as the primary processing units in modern block-based video codecs. When processing high-resolution content such as 4K and 8K video, larger homogeneous areas (such as clear skies or flat walls) appear across more frames. Utilizing a 128x128 maximum superblock size allows AV1 to process large, uniform areas using fewer blocks, which significantly reduces side information, bitstream overhead, and block header metadata compared to older 64x64 architectures.
Recursive Block Partitioning
From the maximum 128x128 level, AV1 provides an extensive recursive partition tree down to blocks as small as 4x4 pixels. While previous standards typically used a 4-way quad-tree split, AV1 expands this to a 10-way partition structure:
- None: The superblock remains unpartitioned.
- Split (Quad-tree): Splits the block into four equal sub-blocks.
- Horizontal Split (2:1): Divides the block horizontally into two equal rectangular blocks.
- Vertical Split (1:2): Divides the block vertically into two equal rectangular blocks.
- Horizontal 4:1 Split: Divides the block into four horizontal strips.
- Vertical 1:4 Split: Divides the block into four vertical strips.
- T-Shaped Partitions: Four asymmetric configurations (horz-A, horz-B, vert-A, vert-B) where one half of the block is split while the other remains intact.
This rich set of partitioning modes gives encoders the flexibility to adapt 128x128 superblocks dynamically to complex object boundaries, fine textures, and continuous motion vectors, yielding substantial bitrate savings at equivalent visual quality.