Video Encoding Intra-Prediction Optimization

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing video encoding methods face limitations in achieving significant speed increases and maintaining high compression ratios due to inadequate optimization of intra-frame block prediction, leading to poorer image quality and larger bit streams.

Innovation Solution

A video encoding method that optimizes intra-prediction by dividing coding units into different sizes, performing initial and fine selections of intra-predicted directions using rate distortion evaluation functions, and selecting optimal brightness and chroma directions to improve encoding speed and compression efficiency.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Manufacturing precision

If intra-frame block prediction is not optimized, then encoding speed is maintained at conventional levels, but compression ratio is greatly reduced and image quality deteriorates

Engineering Contradiction:
Improveimage qualityVSAvoidencoding speed
Core Design Contradiction:
Manufacturing precisionVSProductivity

Solution Approach 1:

The patent segments the intra-frame prediction process into multiple stages: initial mode selection that divides coding units into different sizes, followed by fine selection of prediction directions. This segmentation allows the system to evaluate fewer modes in detail while maintaining accuracy, thus improving both image quality and encoding speed simultaneously

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent performs preliminary action by conducting an initial mode selection that identifies candidate prediction directions before performing the computationally intensive fine selection. This preliminary filtering reduces the number of modes that require detailed evaluation, improving encoding speed without sacrificing image quality

Inventive Principle:
Principle #10Preliminary action

2Quantity of substance

If intra-frame block prediction is not optimized, then encoding process remains simple, but bit stream size increases significantly

Engineering Contradiction:
Improvebit stream sizeVSAvoidprediction method complexity
Core Design Contradiction:
Quantity of substanceVSDevice complexity

Solution Approach 1:

The patent segments the prediction method into distinct phases: initial mode selection with size-based division, and fine selection of directions. This structured approach efficiently reduces bit stream size by ensuring optimal prediction is achieved without requiring excessively complex processing

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent changes parameters dynamically by adjusting the division size of coding units during initial mode selection based on rate-distortion evaluation. This parameter adaptation allows the system to achieve better compression ratios by selecting prediction modes that minimize bit stream size while maintaining manageable computational complexity

Inventive Principle:
Principle #35Parameter changes

3Reliability

If reference pixels have poor image quality, then encoding becomes faster due to simpler processing, but error accumulation increases and overall quality deteriorates

Engineering Contradiction:
Improveerror accumulationVSAvoidencoding efficiency
Core Design Contradiction:
ReliabilityVSProductivity

Solution Approach 1:

The patent performs preliminary action by conducting thorough initial mode selection that evaluates multiple candidate directions before final selection. This preliminary evaluation ensures that high-quality prediction is established early, preventing error accumulation in reference pixels while maintaining encoding efficiency through the subsequent fine selection process

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

The patent implements feedback through rate-distortion evaluation that assesses the quality of prediction modes. This feedback mechanism ensures that only prediction modes achieving sufficient image quality are selected, preventing error accumulation in reference pixels while maintaining efficient encoding through intelligent mode selection

Inventive Principle:
Principle #23Feedback

Data Source

PatentEP3758373B1Video encoding method, electronic device, and storage medium
Publication Date: 2023.09.20 TENCENT TECHNOLOGY (SHENZHEN) CO LTD
  • EP3758373B1 patent drawingFigure 1
  • EP3758373B1 patent drawingFigure 2
  • EP3758373B1 patent drawingFigure 3~4

AI summary

This disclosure provides a video encoding method, including: receiving a video image frame, and acquiring at least one coding tree unit (CTU) of the video image frame; dividing the CTU according to different video prediction unit division rules, to acquire video prediction units of different sizes of the CTU; performing initial selection on intra-predicted brightness prediction directions of the video prediction units of different sizes, to obtain a preset quantity of intra-predicted initially-selected brightness directions of the video prediction units of different sizes; setting intra-predicted optimal brightness directions of the video prediction unit of a current size; performing fine selection on intra-predicted chroma prediction directions of the video prediction units of different sizes, to obtain intra-predicted optimal chroma directions of the video prediction units of different sizes; and performing intra-prediction encoding on a current video image frame according to intra-predicted optimal brightness directions and the intra-predicted optimal chroma directions.