Weighted Prediction for Quantization Matrix Coding Overhead

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

The existing video compression standards, such as H.264/MPEG-4 AVC and HEVC, face challenges in efficiently encoding larger quantization matrices (QMs) due to significant overheads, which cause bandwidth and capacity concerns, especially with the increasing demand for higher video quality and resolution.

Innovation Solution

The method involves using weighted predictive techniques to encode and decode video image data by predicting current QMs from reference QMs and weighting factors, which are then encoded and transmitted as a video bitstream, allowing for lossless or lossy coding and reducing the overhead associated with larger QM sizes.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Productivity

If larger quantization matrices (QMs) are used to enhance coding efficiency and video quality, then coding efficiency and video quality are improved, but the overhead for carrying QMs in the video bitstream increases, causing bandwidth and capacity concerns

Engineering Contradiction:
Improvecoding efficiencyVSAvoidoverhead data size
Core Design Contradiction:
ProductivityVSQuantity of substance

Solution Approach 1:

The patent segments the QM coding process by predicting only the differences between adjacent QM coefficients rather than encoding the entire matrix. This segmentation approach allows efficient representation of QM data by focusing on and encoding only the residual information that needs to be transmitted, significantly reducing overhead while maintaining coding efficiency.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent applies preliminary action by using prediction techniques to estimate QM coefficients before actual encoding. By predicting QM values based on neighboring coefficients and using these predictions to guide the encoding process, the system reduces the amount of data that needs to be transmitted, thereby reducing overhead while preserving coding efficiency.

Inventive Principle:
Principle #10Preliminary action

2Manufacturing precision

If larger quantization matrices (QMs) are used to support higher video resolution and quality, then video resolution and quality are improved, but the bandwidth required to transmit the QM data increases

Engineering Contradiction:
Improvevideo qualityVSAvoidbandwidth
Core Design Contradiction:
Manufacturing precisionVSLength of moving object

Solution Approach 1:

The patent extracts only the essential information needed to represent QM differences by predicting and encoding only the residual coefficients. This extraction approach removes redundant information from the bitstream, allowing high video quality to be maintained while reducing the bandwidth required for transmission.

Inventive Principle:
Principle #2Taking out (Extraction)

Solution Approach 2:

The patent changes the parameter representation by transforming QM data from absolute coefficient values to difference values based on predictions. This parameter transformation allows the same QM information to be represented with fewer bits, thereby maintaining video quality while reducing bandwidth requirements.

Inventive Principle:
Principle #35Parameter changes

3Ease of manufacture

If traditional QM encoding methods are used, then implementation simplicity is maintained, but coding efficiency and bandwidth utilization are insufficient for modern video demands

Engineering Contradiction:
Improveimplementation simplicityVSAvoidcoding efficiency
Core Design Contradiction:
Ease of manufactureVSProductivity

Solution Approach 1:

The patent implements a universal prediction mechanism that can be applied across different QM sizes and video configurations. The same prediction and encoding principles work for various QM block sizes (4x4, 8x8, 16x16, 32x32), providing coding efficiency improvements without requiring complex implementation changes, thus maintaining ease of manufacture while enhancing productivity.

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

Data Source

PatentEP2801194B1Quantization matrix (QM) coding based on weighted prediction
Publication Date: 2021.10.27 HUAWEI TECH CO LTD
  • EP2801194B1 patent drawingFigure 1
  • EP2801194B1 patent drawingFigure 2
  • EP2801194B1 patent drawingFigure 3

AI summary

A method for decoding and reconstructing a QM associated with a picture, wherein the method comprises receiving a video bitstream, acquiring at least one reference QM and at least one weighting factor associated with the reference QM, wherein the reference QM is a previously decoded QM, and computing a predicted matrix using the reference QM and the weighting factor associated with the reference QM, wherein the predicted matrix is used to reconstruct the QM. In another embodiment, a method for encoding a QM associated with a picture, wherein the method comprises acquiring a first QM reference and a second QM reference, acquiring a first weighting factor that corresponds to the first QM reference and a second weighting factor that corresponds to the second QM reference, obtaining the predicted QM using the first QM reference, the second QM reference, the first weighting factor, and the second weighting factor.