Patent Pool Technical Arguments Against AV1

The Alliance for Open Media designed the AV1 video codec to deliver high-efficiency video compression under a royalty-free license, yet patent pools such as Sisvel have formed licensing programs claiming that AV1 infringes upon hundreds of essential patents. Patent pools challenge AV1’s royalty-free status by asserting that modern video compression relies on a universally shared technical foundation, making it virtually impossible to achieve high compression ratios without using patented technologies. This article breaks down the primary technical arguments and specific codec subsystems where patent pools allege infringement.

Structural Reliance on the Hybrid Video Coding Model

Patent pools argue that AV1 does not fundamentally diverge from the foundational architecture used in standardized codecs like AVC (H.264), HEVC (H.265), and VVC (H.266). AV1 operates as a traditional hybrid block-based video codec using a sequence of:

• Spatial and temporal prediction
• Residual transform and quantization
• In-loop filtering
• Entropy coding

Because AV1 uses this classical pipeline, patent pools contend that baseline operational principles—such as residual data signaling, syntax layout, and reconstruction pipelines—overlap directly with broad, foundational patent claims protecting the mechanics of hybrid video coding.

Advanced Block Partitioning and Superblock Hierarchies

A major source of coding efficiency in AV1 is its flexible partitioning framework, which starts with a 128x128 or 64x64 superblock and recursively divides it into smaller blocks down to 4x4. AV1 supports recursive quad-tree, horizontal/vertical splits, and 1:4/4:1 non-square rectangular partitions.

Patent pools assert that these partitioning schemes infringe upon patent portfolios developed during the HEVC standardization process. The core technical arguments center on:

• Signaling mechanisms used to communicate block partition trees within the bitstream.
• Boundary-handling processes for blocks overlapping picture boundaries.
• Partition size restrictions and contextual decision logic applied during recursive splitting.

Intra and Inter Prediction Mechanisms

Patent licensors argue that AV1’s high compression gains rely on predictive algorithms that closely mimic techniques patented during the development of ISO/IEC and ITU-T standards.

Directional Intra Prediction

AV1 expands directional intra prediction to 56 directional angles along with angle offsets to model complex textures. Patent holders claim infringement regarding the specific mathematical interpolation methods used to generate reference samples at fractional pixel positions and the methods used to filter reference boundaries prior to intra estimation.

Advanced Inter Prediction

To minimize motion vector overhead, AV1 employs sophisticated temporal prediction tools that patent pools heavily scrutinize:

• Compound and Dual Prediction: Combining multiple motion vectors with spatial weighting to predict a single block.
• Overlapped Block Motion Compensation (OBMC): Blending predictions using neighboring motion vectors to reduce boundary artifacts.
• Warped and Affine Motion Compensation: Modeling non-translational motion like zooming, rotation, and shear.

Patent pools hold extensive portfolios covering the mathematical derivation, rounding operations, and reference frame index signaling methods necessary to perform these advanced motion compensation operations.

In-Loop Filtering Subsystems

AV1 employs a three-stage in-loop filtering process to eliminate compression artifacts before using reconstructed frames as reference images: the Deblocking Filter, the Constrained Directional Enhancement Filter (CDEF), and the Loop Restoration Filter (Wiener and self-guided restoration).

Patent pools argue that:

• CDEF Edge Detection: The directional search algorithm within CDEF analyzes local pixel variance and determines edge direction using mathematical formulations patented by legacy codec contributors.
• Filter Parameter Signaling: The syntax and conditional criteria used to enable, disable, and transmit filter coefficients across tile- and frame-level boundaries infringe on patents regarding adaptive filtering signaling.

Entropy Coding and Context Adaptation

AV1 utilizes a multi-symbol arithmetic coder known as the Daala entropy coder, designed to bypass the binary arithmetic coding patents covering CABAC (Context-Adaptive Binary Arithmetic Coding).

Despite architectural differences from CABAC, patent pools argue that AV1’s context modeling—specifically how the codec updates probabilities, determines context neighborhoods from neighboring transform blocks, and dynamically tracks symbol probabilities across frames—infringes on broader intellectual property covering adaptive entropy estimation and probability-state transitions.

Broad Functional Claim Scope

Beyond specific algorithm-level overlaps, patent pools leverage broadly drafted functional claims. These patents describe the technical objective or high-level functional step (such as identifying a motion vector candidate from a spatial neighbor under defined conditions) rather than a rigid, single-codec implementation. Patent licensors argue that by implementing features essential for competitive 4K and 8K compression efficiency, AV1 inevitably executes operations claimed within their extensive patent catalogs.