History-Based Motion Vector Predictor Selection Order

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Current video coding techniques face challenges in efficiently compressing large video data while maintaining high quality, particularly due to the limitations of existing motion vector prediction methods which do not effectively utilize historical motion vector information for improved coding efficiency.

Innovation Solution

The implementation of a history-based motion vector prediction method where motion vector predictors are obtained from a list of previously decoded vectors, with a First-In-First-Out (FIFO) table that updates and selects candidates based on their recency, allowing for more effective selection in inter-prediction modes like merge and Advanced Motion Vector Prediction (AMVP) modes.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Productivity

If traditional motion vector prediction methods are used, then the coding process is simple, but coding efficiency is insufficient and bit rate reduction is limited

Engineering Contradiction:
Improvecoding efficiencyVSAvoidprediction process complexity
Core Design Contradiction:
ProductivityVSDevice complexity

Solution Approach 1:

The patent applies preliminary action by pre-sorting HMVP candidates in ascending order of their reference picture list indices before they are added to the candidate list. This pre-arrangement ensures that when candidates are subsequently selected for prediction, those with smaller reference picture indices are considered first, eliminating the need for complex real-time sorting or selection logic during the prediction process itself.

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

The patent implements dynamics by adaptively adjusting the selection order of HMVP candidates based on their reference picture list indices. Instead of using a fixed selection order, the system dynamically determines which candidate to select first by comparing their reference picture indices, allowing the prediction process to adapt to different coding scenarios and reference picture configurations.

Inventive Principle:
Principle #15Dynamics

2Measurement precision

If more historical motion vector candidates are stored, then prediction accuracy improves, but memory usage and processing overhead increase

Engineering Contradiction:
Improvemotion vector prediction accuracyVSAvoidmemory storage requirement
Core Design Contradiction:
Measurement precisionVSQuantity of substance

Solution Approach 1:

The patent applies the taking out principle by selectively extracting and prioritizing HMVP candidates with smaller reference picture list indices. Instead of treating all HMVP candidates equally or storing all possible candidates, the system extracts the most relevant candidates based on their reference picture index values, thereby reducing the effective number of candidates that need to be maintained and processed while still achieving accurate prediction.

Inventive Principle:
Principle #2Taking out (Extraction)

Solution Approach 2:

The patent implements local quality by assigning different priorities to different HMVP candidates based on their specific reference picture list indices. Candidates with smaller indices are given higher priority and are selected first, while candidates with larger indices are considered later or may be discarded if the candidate list is full. This differentiated treatment optimizes prediction accuracy by focusing resources on the most relevant candidates.

Inventive Principle:
Principle #3Local quality

Data Source

PatentEP3864841B1Improvements on history-based motion vector predictor
Publication Date: 2025.01.01 QUALCOMM INC
  • EP3864841B1 patent drawingFigure 1
  • EP3864841B1 patent drawingFigure 2A~2B
  • EP3864841B1 patent drawingFigure 3A~3B

AI summary

Systems and techniques for processing video data include a history-based motion vector predictor (HMVP) table with two or more HMVP candidates, where a first entry of the HMVP table associated with a first index value includes a first HMVP candidate and a second entry of the HMVP table associated with a second index value includes a second HMVP candidate, the first index value being lower than the second index value. For an advanced motion vector prediction (AMVP) candidate list, one or more HMVP candidates from the HMVP table are selected in a reverse order, where the second HMVP candidate is selected before the first HMVP candidate according to the reverse order. The selected one or more HMVP candidates are added to the AMVP candidate list. The AMVP candidate list can be used to perform AMVP for the one or more blocks of video data.