Quantizing Matrix Generation for Video Coding Efficiency

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing video coding methods face a drop in coding efficiency due to the signaling of quantizing matrices, and the process cost for generating these matrices is high, especially in future video coding (FVC) with varied block sizes and shapes.

Innovation Solution

An image processing device and method that efficiently generate or signal quantizing matrices by decoding scaling list data to create a first quantizing matrix, and then generating a second quantizing matrix for a transform block with zeroed high-frequency components by referencing only a partial matrix of the first quantizing matrix.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Adaptability or versatility

If all quantizing matrices for different TU sizes are signaled in the coded bit stream, then the diversity of quantizing matrices is improved, but the overhead increases and coding efficiency decreases

Engineering Contradiction:
Improvediversity of quantizing matricesVSAvoidoverhead in coded bit stream
Core Design Contradiction:
Adaptability or versatilityVSLoss of information

Solution Approach 1:

The patent segments the quantizing matrix signaling process by dividing matrices into different size categories (e.g., 4×4, 8×8, 16×16, 32×32, 64×64, 128×128). Instead of signaling all matrices individually, it signals only representative matrices for each size category, thereby reducing overhead while maintaining the ability to handle diverse block sizes through the segmented structure.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent extracts and signals only the essential quantizing matrices that are most frequently used or most impactful for coding efficiency. By selecting a subset of representative matrices from the full set of possible matrices across different TU sizes, it reduces the signaling overhead while preserving the necessary diversity for effective compression.

Inventive Principle:
Principle #2Taking out (Extraction)

2Loss of information

If multiple quantizing matrices are generated from a single signaled matrix, then the overhead is reduced, but the process cost for generation increases

Engineering Contradiction:
Improveoverhead in coded bit streamVSAvoidprocess cost for generation
Core Design Contradiction:
Loss of informationVSDevice complexity

Solution Approach 1:

The patent performs preliminary actions by pre-defining the generation rules and relationships between quantizing matrices of different sizes. The decoder is pre-programmed with the knowledge of how to generate smaller or larger matrices from signaled representative matrices, thereby reducing the computational burden during actual decoding operations.

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

The patent uses copying mechanisms where quantizing matrices for certain TU sizes are generated by copying and adapting from signaled representative matrices. For example, matrices for smaller block sizes can be generated by copying relevant portions or applying scaling factors to the signaled matrix, thereby reducing generation complexity while maintaining consistency.

Inventive Principle:
Principle #26Copying

3Adaptability or versatility

If quantizing matrices are generated for all possible TU sizes including 128×128 and non-square TUs, then the adaptability to FVC is improved, but the process cost and hardware resource occupation increase

Engineering Contradiction:
Improvesupport for expanded TU sizesVSAvoidhardware resources and processing delays
Core Design Contradiction:
Adaptability or versatilityVSDevice complexity

Solution Approach 1:

The patent implements universality by designing a quantizing matrix generation system that can handle multiple TU sizes (including the expanded 128×128 and non-square sizes) using a unified approach. A single signaled representative matrix can serve as the basis for generating matrices for various other sizes through systematic scaling and adaptation, thereby reducing the need for separate dedicated matrices for each size and reducing overall device complexity.

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

Data Source

PatentEP4550782A1Image processing devices and image processing methods
Publication Date: 2025.05.07 SONY GROUP CORP
  • EP4550782A1 patent drawingFigure 1
  • EP4550782A1 patent drawingFigure 2
  • EP4550782A1 patent drawingFigure 3A~3B

AI summary

[Problem] An improved system for efficiently generating or signaling quantized matrices is provided. [Solution] An image processing device is provided. The image processing device includes a decoding unit that decodes scaling list data to generate a first quantizing matrix of a first size, a generating unit that generates a second quantizing matrix for a transform block of a second size to which zeroing of a high-frequency component is applied, by referring to only a partial matrix of the first quantizing matrix generated by the decoding unit, and an inverse quantizing unit that inversely quantizes a quantized transform coefficient of the transform block of the second size, using the second quantizing matrix generated by the generating unit.