Inter-Prediction Motion Vector Refinement for Video Encoding

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Current video compression techniques, such as H.264/AVC and HEVC, face challenges in improving encoding efficiency and image quality due to increasing picture size, resolution, and frame rate, particularly in inter-prediction processes where motion vector refinement is complex and computationally intensive.

Innovation Solution

A coding tool that adjusts motion vector corrections using high-level syntax, enabling simplified integer sample search and fractional sample refinement operations to enhance inter-prediction compression performance by decoding high-level syntax elements and generating refined motion vectors for improved prediction accuracy.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Manufacturing precision

If motion estimation and compensation are used to improve video quality, then video quality is improved, but computational complexity increases

Engineering Contradiction:
Improvevideo qualityVSAvoidcomputational complexity
Core Design Contradiction:
Manufacturing precisionVSDevice complexity

Solution Approach 1:

The patent divides the prediction process into multiple refinement stages (first refinement, second refinement, third refinement) where each stage processes only the difference between previous predictions and actual values. This segmentation allows the system to achieve high prediction accuracy through incremental improvements rather than requiring one complex exhaustive calculation, thereby reducing overall computational complexity while maintaining video quality.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent performs preliminary motion estimation using a first motion estimation method to obtain initial motion vectors before applying refinement methods. This preliminary action provides a baseline prediction that is then progressively improved, avoiding the need to perform exhaustive optimization calculations from scratch and reducing the total computational burden while maintaining accuracy.

Inventive Principle:
Principle #10Preliminary action

2Measurement precision

If multiple motion estimation methods are applied to improve prediction accuracy, then prediction accuracy is improved, but processing time increases

Engineering Contradiction:
Improveprediction accuracyVSAvoidprocessing time
Core Design Contradiction:
Measurement precisionVSLoss of time

Solution Approach 1:

The patent applies motion estimation and compensation in a periodic iterative manner across multiple refinement stages. Each stage performs a limited number of iterations focused on specific refinement aspects, then moves to the next stage. This periodic action allows the system to achieve high accuracy through structured incremental improvements rather than continuous exhaustive searching, reducing overall processing time.

Inventive Principle:
Principle #19Periodic action

Solution Approach 2:

The patent performs preliminary motion estimation using a first motion estimation method to obtain initial motion vectors before applying refinement methods. This preliminary action provides a baseline prediction that is then progressively improved, avoiding the need to perform exhaustive optimization calculations from scratch and reducing the total computational burden while maintaining accuracy.

Inventive Principle:
Principle #10Preliminary action

3Manufacturing precision

If detailed motion vector refinement is performed to improve encoding precision, then encoding precision is improved, but computational load increases

Engineering Contradiction:
Improveencoding precisionVSAvoidcomputational load
Core Design Contradiction:
Manufacturing precisionVSPower

Solution Approach 1:

The patent applies different refinement strategies to different regions or aspects of the motion estimation process. The first refinement method focuses on sub-pixel accuracy, the second on directional accuracy, and the third on boundary accuracy. This local quality approach ensures that computational resources are concentrated on the most critical aspects of prediction accuracy for each specific case, improving encoding precision without uniformly increasing computational load across all scenarios.

Inventive Principle:
Principle #3Local quality

Data Source

PatentEP3993423A1Method and device for encoding and decoding videoby using inter-prediction
Publication Date: 2022.05.04 HYUNDAI MOTOR CO LTD
  • EP3993423A1 patent drawingFigure 1
  • EP3993423A1 patent drawingFigure 2
  • EP3993423A1 patent drawingFigure 3A~3B

AI summary

A coding tool is provided to improve the compression performance of inter prediction and is used at the encoder/decoder side to adjust the correction of a motion vector based on a high level syntax. In addition, a method is provided for simply performing an integer sample search step of searching for an integer offset and a fractional sample refinement step of searching for a sub-pixel offset in relation to motion vector refinement among coding tools.