AV2 Quantizer Design for Lossless and Near-Lossless Compression

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

The existing AV1 and proposed AV2 quantization methods result in higher bitrates for lossy coding modes compared to lossless modes, leading to inefficiencies in video compression.

Innovation Solution

A modified quantization approach for AV2 that adjusts quantization step sizes based on specific conditions, allowing for efficient lossless or near-lossless compression by modifying quantization indices and step sizes using unified lookup tables and additional sequence-level syntax parameters.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Loss of energy

If existing AV1 and proposed AV2 quantization methods are used, then video compression is achieved, but bitrate increases for lossy coding modes compared to lossless modes

Engineering Contradiction:
Improvebitrate efficiencyVSAvoidcompression efficiency
Core Design Contradiction:
Loss of energyVSProductivity

Solution Approach 1:

The patent modifies quantization parameters dynamically based on the coding mode (lossless vs. lossy). Specifically, it adjusts the quantization step size and quantization index values differently for lossless and lossy modes, allowing optimal bitrate efficiency for each mode without compromise

Inventive Principle:
Principle #35Parameter changes

Solution Approach 2:

The quantization process is made adaptive and dynamic by selecting different quantization strategies based on the operating mode. The system transitions between fixed quantization (lossless) and variable quantization (lossy) approaches, optimizing performance for each scenario

Inventive Principle:
Principle #15Dynamics

2Loss of energy

If quantization index values are reduced for lossless mode, then bitrate is reduced, but compression efficiency decreases

Engineering Contradiction:
ImprovebitrateVSAvoidcompression efficiency
Core Design Contradiction:
Loss of energyVSLoss of information

Solution Approach 1:

The patent applies different quantization quality levels to different coefficient types (DC vs. AC, luma vs. chroma) within the same coding mode. This localized quality adjustment ensures that critical components maintain high quality while less critical components use coarser quantization, optimizing the bitrate-efficiency tradeoff

Inventive Principle:
Principle #3Local quality

3Reliability

If separate quantization methods are used for lossless and lossy modes, then mode-specific optimization is achieved, but device complexity increases

Engineering Contradiction:
Improvemode-specific performanceVSAvoidquantization system complexity
Core Design Contradiction:
ReliabilityVSDevice complexity

Solution Approach 1:

The patent uses a unified quantization framework that can operate in multiple modes (lossless and lossy) by adjusting parameters. The same quantization infrastructure serves both modes, reducing overall system complexity while maintaining mode-specific optimization through parameter adjustment rather than separate implementations

Inventive Principle:
Principle #6Universality (Multi-functionality)

Data Source

PatentEP4008110B1Quantizer for lossless & near-lossless compression
Publication Date: 2025.09.17 TENCENT AMERICA LLC
  • EP4008110B1 patent drawingFigure 1
  • EP4008110B1 patent drawingFigure 2
  • EP4008110B1 patent drawingFigure 3

AI summary

A system includes code including obtaining code to obtain a first syntax element that indicates a first quantization index value; at least one second syntax element that indicates an offset value, a second quantization index value for another coefficient by combining the first quantization index value and the offset value to obtain a combined value, and modifying the combined value to be a predetermined minimum value as the second quantization index value, fourth obtaining code to obtain a quantization step size that corresponds to the second quantization index value; and determining code to determine a mode in which the coded image is to be decoded based on determining whether the first quantization index value is equal to a quantization index value associated with lossless coding, and based on determining whether the offset value is less than or equal to the quantization index value associated with the lossless coding.