Bidirectional Intra-Prediction Video Coding

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing video coding intra-frame prediction methods have low prediction accuracy due to the distance between reference prediction pixels and the current block, leading to reduced coding efficiency.

Innovation Solution

A video coding method that performs bidirectional prediction, including intra-frame left and upper direction prediction and intra-frame right and lower direction prediction, and weights these predictions to improve accuracy, especially when the distance between the current block and reference pixels is closer.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Device complexity

If linear prediction from left and upper adjacent rows is used, then the prediction process is simple, but the prediction accuracy deteriorates for pixels farther from reference pixels

Engineering Contradiction:
Improveprediction process complexityVSAvoidprediction accuracy
Core Design Contradiction:
Device complexityVSMeasurement precision

Solution Approach 1:

The current block is divided into multiple sub-blocks, and different prediction directions are applied to different sub-blocks. Specifically, the block is segmented into regions that use left-reference predictions and regions that use upper-reference predictions, allowing each region to utilize the most relevant reference pixels for accurate prediction.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent transitions from unidirectional prediction (only left or only upper) to bidirectional prediction by incorporating both horizontal and vertical reference dimensions. This dimensional expansion allows the prediction to leverage reference pixels from both left and upper directions simultaneously, improving accuracy for pixels distant from single-direction references.

Inventive Principle:
Principle #17Another dimension (Dimensionality change)

2Device complexity

If reference pixels from only left and upper positions are used, then the reference source is limited, but the prediction correlation decreases for right and lower pixels

Engineering Contradiction:
Improvereference source configurationVSAvoidprediction correlation
Core Design Contradiction:
Device complexityVSReliability

Solution Approach 1:

The patent applies prediction in multiple directions including not only the traditional left and upper directions but also incorporates right and lower reference predictions. By inverting the conventional single-direction approach and utilizing opposite directions (right and lower), the system improves prediction correlation for pixels that are distant from the original reference positions.

Inventive Principle:
Principle #13The other way round (Inversion)

Data Source

PatentEP3509305B1Intra-prediction video coding method and device
Publication Date: 2021.06.16 BEIJING QIYI CENTURY SCI & TECH CO LTD
  • EP3509305B1 patent drawingFigure 1~2
  • EP3509305B1 patent drawingFigure 3~4
  • EP3509305B1 patent drawingFigure 5~6

AI summary

The present invention provides a video coding intra-frame prediction method and a video coding intra-frame prediction device, wherein, the method comprises performing a bidirectional prediction of an intra-frame left and upper direction prediction and an intra-frame right and lower direction prediction on the current coding block to obtain a prediction result; weighting the prediction result obtained. Since the distance between the current coding block and the reference prediction pixel is closer, thus the prediction correlation is high, thus it could achieve the high accuracy prediction of the intra-frame prediction during the process of video coding and thus could improve the coding efficiency.