Ithy Logo

AV1 Video Coding Techniques

Advanced technologies shaping next-gen video compression

av1 codec illustration

Key Takeaways

  • High Compression Efficiency: Achieved through advanced block partitioning and transform coding.
  • Versatile Prediction Models: Including intra, inter, and screen content coding.
  • Enhanced Visual Quality: Via sophisticated loop filters and motion compensation.

Block Partitioning and Coding Structures

Flexible Partitioning Schemes

AV1 employs a highly adaptable block partitioning system known as Flexible Quadtree + Binary Tree Partitioning (QTBTT). This system allows video frames to be divided into superblocks of varying sizes, ranging from 4x4 to 128x128 pixels. These superblocks can be recursively split into smaller blocks using a combination of quadtree and binary tree partitioning, enabling precise modeling of diverse spatial characteristics within a frame (arXiv).

Superblocks and Partitioning Flexibility

Superblocks, which are the largest block units in AV1, can be partitioned into smaller blocks through various patterns such as two-way, four-way, T-shaped, horizontal, and vertical splits. This flexibility allows the codec to efficiently encode both flat regions and areas with intricate textures or high detail, optimizing compression efficiency and maintaining high visual fidelity (ImageKit).


Prediction Techniques

Intra-Frame Prediction

Intra-frame prediction in AV1 leverages multiple prediction modes, including directional, non-directional, and chroma-from-luma predictions. With up to 56 intra-prediction modes, AV1 can accurately estimate pixel values within a single frame based on neighboring pixels, significantly reducing redundancy and enhancing compression efficiency (Wikipedia).

Inter-Frame Prediction

AV1's inter-frame prediction utilizes advanced motion compensation techniques with multi-directional motion vectors. It supports both simple and compound predictions, allowing the combination of multiple reference frames to better predict the current frame's content. This includes the use of warped motion compensation and global motion models to handle complex scene transformations, improving the accuracy of motion prediction and reducing bitrate requirements (ImageKit).

Screen Content Coding (SCC)

For content with sharp edges, text, and repetitive elements, AV1 incorporates Screen Content Coding (SCC) features such as Intra Block Copy (IBC) and Palette Coding. Palette Coding efficiently encodes areas with limited color palettes by using palette indices instead of full color values, substantially reducing the required bitstream size (Visionular).


Transform and Quantization Coding

Multi-Type Transforms (MTT)

AV1 employs a variety of transform types, including Discrete Cosine Transform (DCT) and Discrete Sine Transform (DST), to decorrelate image data effectively. This enables efficient encoding of both smooth areas and abrupt edges by transforming spatial domain data into the frequency domain, facilitating more effective compression (Medium).

Adaptive Quantization

AV1 utilizes adaptive quantization techniques that dynamically adjust quantization parameters based on the complexity of the content. This allows for more aggressive compression in less detailed areas while preserving quality in high-detail regions. Segment-Based Adaptive Quantization (AQ) and Delta Q provide per-block quantizer adjustments, ensuring optimal bitrate distribution and maintaining visual fidelity where it matters most (Flussonic).


Entropy Coding

Context-Adaptive Binary Arithmetic Coding (CABAC)

AV1 implements Context-Adaptive Binary Arithmetic Coding (CABAC), a sophisticated entropy coding system that adapts encoding symbols based on the local probability of occurrences. This approach enhances coding efficiency by allowing multiple bits to be encoded simultaneously and improving the probability estimation based on surrounding data (arXiv).

Multi-Symbol Arithmetic Coding

Unlike traditional binary arithmetic coding used in earlier codecs like VP9, AV1's multi-symbol arithmetic coding enables higher compression ratios by encoding multiple symbols in a single operation. This advancement reduces the bitstream size further while maintaining high decoding efficiency (ImageKit).

Symbol Pruning and Level-Map Coding

AV1 employs symbol pruning techniques to select the most probable symbols for representation, effectively reducing coding redundancies. Additionally, level-map based coefficient coding allows for further lossless compression by efficiently encoding coefficients based on their levels (Cambridge).


Motion Compensation and Frame Referencing

Advanced Motion Compensation

AV1 introduces sophisticated motion compensation tools such as Warped Motion and Global Motion Models. These tools go beyond simple translational predictions to handle rotation and perspective transformations, allowing for more accurate motion prediction in complex scenes. Overlapped Block Motion Compensation (OBMC) blends predictions with neighboring blocks, reducing blockiness and improving visual quality (arXiv).

Frame Referencing

The codec expands frame referencing capabilities by utilizing seven out of eight available frames in the decoded frame buffer, including LAST, GOLDEN, ALTREF, LAST2, and LAST3 frames. This extensive referencing enables AV1 to analyze motion and predict pixel changes more accurately, enhancing the codec's ability to handle intricate motion and temporal dependencies (Bunny.net).

Switch Frames

Switch Frames in AV1 utilize already-decoded reference frames from higher-resolution versions of the same video. This feature facilitates smooth resolution switching in adaptive bitrate streaming, allowing efficient bitrate adaptation without the need for full keyframes at the beginning of each video segment, thereby maintaining video quality during streaming adjustments (ImageKit).


Loop Filters and Post-Processing

Deblocking and Directional Filters

AV1 incorporates in-loop deblocking filters that reduce blocking artifacts between encoded blocks by smoothing block boundaries. Additionally, Directional Enhancement Filters and Constrained Directional Enhancement Filters (CDEF) further refine the image quality by addressing diagonal and textured content, enhancing overall visual fidelity (JM Valin PDF).

Self-Guided Restoration Filters

To refine decoded frames, AV1 employs Self-Guided Restoration Filters, including Bilateral Solver and Wiener filters. These filters address fine details and noise, effectively enhancing image quality without significantly increasing decoding complexity (Visionular).

Loop Restoration and Super-Resolution

Loop Restoration Filters and Frame Super-Resolution techniques in AV1 work within the encoding/decoding loop to further enhance image details. These techniques improve the sharpness and clarity of the video, especially in high-resolution content, ensuring that visual quality is maintained even after aggressive compression (MakeUseOf).


Adaptive and Scalable Tools

Rate Control and Adaptive Quantization

AV1's rate control mechanisms ensure that video bitrate remains within desired limits while maintaining quality. Adaptive Quantization dynamically adjusts the quantization parameters based on the complexity of different content regions, allocating more bits to complex areas and fewer bits to simpler ones. This balance optimizes the overall bitrate distribution and enhances compression efficiency (Flussonic).

Segment-Based Adaptive Quantization (AQ) and Delta Q

Segment-Based Adaptive Quantization (AQ) allows different regions of a frame to be quantized differently based on their complexity. Delta Q enables per-block quantizer adjustments, providing greater flexibility in maintaining visual fidelity in critical areas while saving bits in less important regions (Bunny.net).

Toolset Extensibility and Future-Proofing

Designed with extensibility in mind, AV1 can incorporate new features and optimizations as technology evolves. This future-proofing ensures that AV1 remains adaptable to emerging video formats and display technologies, securing its relevance in the ever-changing landscape of video coding standards (Wikipedia).


Scalability and Hardware Considerations

Parallel Processing and Multi-threading

AV1 is optimized for modern multi-core processors, utilizing tile-based parallelism to divide frames into independently decodable tiles. This allows for efficient parallel processing during both encoding and decoding, significantly enhancing performance on multi-core systems (IEEE).

Vectorization and Hardware Optimization

Incorporating Single Instruction, Multiple Data (SIMD) and other vectorization techniques, AV1 ensures better utilization of modern hardware capabilities. These optimizations enable faster processing speeds and lower power consumption, making AV1 suitable for a wide range of devices, from smartphones to high-end servers (Flussonic).

Hardware Feasibility and Scalability

AV1 is designed with hardware feasibility and scalability in mind, supporting various encoding profiles to match different hardware capabilities. This ensures that AV1 can be efficiently implemented across a diverse array of hardware platforms, facilitating widespread adoption and integration (Bunny.net).


Enhanced Coding Modes

Palette Mode

Palette Mode in AV1 is designed for encoding areas with a limited number of colors, such as logos or text overlays. By representing colors using palette indices instead of full color values, Palette Mode significantly reduces the bitstream size while maintaining color fidelity. This is particularly beneficial for screen content and animated graphics (Visionular).

Intra Block Copy (IntraBC)

Intra Block Copy (IntraBC) allows blocks within the same frame to reference previously decoded areas, enabling self-referencing within the frame. This technique is especially effective for encoding screen content, where repetitive patterns and sharp edges are common, further enhancing compression efficiency (Visionular).

Switch Frames and Alternate Reference Frames

Switch Frames and Alternate Reference Frames (ALTREF) provide additional flexibility in frame referencing. ALTREF frames are generated through temporal filtering across multiple frames, improving compression for low-motion and static regions. Switch Frames utilize higher-resolution references to facilitate smooth resolution switching in adaptive streaming scenarios, enhancing the overall adaptability of the codec (MakeUseOf).


Summary of AV1's Efficiency Gains

AV1 outperforms its predecessors like VP9 and competitors such as HEVC by offering an estimated bitrate reduction of 30-40% at equivalent quality levels (Google Research Paper). This significant improvement is achieved through a comprehensive suite of advanced techniques, including flexible block partitioning, versatile prediction models, sophisticated entropy coding, and enhanced post-processing filters. These innovations not only provide higher compression efficiency but also maintain high visual quality, making AV1 a leading choice for modern video streaming, broadcasting applications, and high-resolution content delivery.


Last updated January 10, 2025
Search Again