Quantization Parameter Set Generation for Video Coding Efficiency

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Conventional image or video coding schemes face inefficiencies in coding quantization scaling matrices and offset matrices, leading to increased data volume and decreased coding efficiency, especially when multiple matrices are required for different pictures.

Innovation Solution

A method to generate a new quantization parameter set from an existing one, allowing the update parameter to be coded instead of the entire new matrix, thereby reducing the amount of data needed for coding and improving efficiency.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Manufacturing precision

If multiple quantization scaling matrices and offset matrices are coded in picture headers for different pictures, then image quality is improved through picture-specific quantization, but the amount of coding data increases significantly

Engineering Contradiction:
Improveimage qualityVSAvoidcoding data volume
Core Design Contradiction:
Manufacturing precisionVSQuantity of substance

Solution Approach 1:

The patent applies parameter changes by transitioning from coding complete quantization matrices to coding only differential update parameters. The quantization matrices are updated by applying delta values (update parameters) to the base matrices, thereby changing the representation from full-parameter to differential-parameter form. This reduces the data volume while preserving the ability to achieve picture-specific quantization for improved image quality.

Inventive Principle:
Principle #35Parameter changes

Solution Approach 2:

The patent uses copying by maintaining a base quantization matrix that is reused across multiple pictures, and only copying the differential update parameters for each picture. Instead of transmitting complete matrices for each picture, the system copies the base matrix and applies picture-specific updates, thereby reducing redundant data transmission while maintaining picture-specific optimization capability.

Inventive Principle:
Principle #26Copying

2Productivity

If quantization matrices are updated frequently for each picture, then coding efficiency is improved through adaptive quantization, but the complexity of the coding system increases

Engineering Contradiction:
Improvecoding efficiencyVSAvoidcoding system complexity
Core Design Contradiction:
ProductivityVSDevice complexity

Solution Approach 1:

The patent applies segmentation by separating the quantization matrix into a base matrix component and a differential update component. The base matrix is coded once and reused, while the differential updates are coded separately for each picture. This segmentation allows the system to achieve adaptive quantization efficiency without requiring complete matrix re-coding, thereby improving coding efficiency while managing system complexity through modular processing.

Inventive Principle:
Principle #1Segmentation

3Measurement precision

If the entire quantization matrix is re-coded for each picture, then accuracy is maintained, but the number of bits required increases

Engineering Contradiction:
Improvequantization accuracyVSAvoidbit rate
Core Design Contradiction:
Measurement precisionVSLoss of information

Solution Approach 1:

The patent extracts only the essential differential information needed to update the quantization matrix from the complete matrix data. By taking out only the update parameters (differences from the base matrix) rather than re-coding the entire matrix, the system maintains quantization accuracy for each picture while significantly reducing the bit rate required for transmission.

Inventive Principle:
Principle #2Taking out (Extraction)

Data Source

PatentEP2495973B1Decoding method, decoder apparatus, encoding method, and encoder apparatus
Publication Date: 2019.02.27 SUN PATENT TRUST
  • EP2495973B1 patent drawingFigure 1
  • EP2495973B1 patent drawingFigure 2
  • EP2495973B1 patent drawingFigure 3

AI summary

A decoding method includes: obtaining a plurality of quantization parameter sets from a header of a coded stream (S400); parsing an identifier from a header of a coded picture included in the coded stream (S402); selecting at least one quantization parameter set from among the plurality of quantization parameter sets, based on the parsed identifier (S404); determining whether a flag parsed from the header of the coded picture has a predetermined value (S408); generating a new quantization matrix from another quantization matrix when the flag has the predetermined value (S410); decoding the coded picture by inversely quantizing the coded picture using the generated new quantization matrix (S412); and decoding the coded picture by inversely quantizing the coded picture using a quantization matrix included in the selected quantization parameter set when the flag does not have the predetermined value (S414).