Multi-Motion Model Video Coding Using HMVP Look-Up Tables
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current video coding standards, such as HEVC and AVS3, face challenges in efficiently utilizing motion information from adjacent and non-adjacent blocks for improved motion vector prediction, leading to suboptimal coding efficiency and increased bandwidth demand.
Innovation Solution
The implementation of History-Based Motion Vector Prediction (HMVP) techniques using look-up tables (LUTs) to store and predict motion information from previously coded blocks, allowing for the inclusion of multiple motion candidates and adaptive searching orders based on motion models, enhances motion vector prediction and coding efficiency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If motion information from previously coded blocks is utilized using HMVP techniques, then coding efficiency is improved and bitrate is reduced, but device complexity increases due to look-up table storage and management
Solution Approach 1:
The patent applies preliminary action by pre-storing motion information from previously coded blocks in look-up tables (HMVP tables) before the actual motion vector prediction process. This allows the decoder to quickly access historical motion data without performing complex real-time searches, thereby improving coding efficiency while managing complexity through advance preparation of motion candidates.
Solution Approach 2:
The patent uses copying by creating simplified copies of motion information from previously coded blocks and storing them in HMVP look-up tables. Instead of storing complete motion data structures, the system copies essential motion vectors and parameters into a compact table format, reducing memory requirements and access complexity while maintaining prediction accuracy.
2Measurement precision
If multiple motion candidates from the same motion model are included, then prediction accuracy is improved, but processing time increases due to expanded search space
Solution Approach 1:
The patent applies local quality by treating different motion model candidates differently in the selection process. Instead of uniformly processing all motion candidates, the system identifies candidates from the same motion model and gives them priority or weighted consideration. This localized optimization improves prediction accuracy for blocks with consistent motion characteristics while avoiding unnecessary processing of diverse motion models that would increase computation time.
Solution Approach 2:
The patent uses partial action by selectively including multiple motion candidates from the same motion model only when beneficial, rather than always processing all possible candidates. The system evaluates whether expanding the candidate set from the same motion model will improve prediction accuracy and applies this expansion selectively, thereby improving accuracy when needed while controlling processing time through conditional application.
3Productivity
If adaptive searching orders based on motion models are implemented, then coding efficiency is improved, but algorithm complexity increases
Solution Approach 1:
The patent applies dynamics by implementing adaptive searching orders that change based on the detected motion model characteristics. Instead of using a fixed search order for all blocks, the system dynamically adjusts the search sequence according to the motion model identified in previously coded blocks. This dynamic adaptation improves coding efficiency by prioritizing relevant motion candidates while managing algorithm complexity through rule-based adjustments rather than complex optimization algorithms.
Data Source
AI summary
Methods, systems and devices for multi-motion model based video coding and decoding are described. An exemplary method for video processing includes determining, for a video block, a candidate for decoding using, for one or more target motion models from a number of motion models, one or more motion candidates from corresponding non-adjacent spatial or temporal blocks or motion information derived from previously coded blocks based on their associated motion models, and performing further processing of the video block using the candidate.