HMVP Motion Inheritance for Lower-Complexity Video Decoding

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing video coding technologies face inefficiencies in updating and managing history-based motion vector prediction (HMVP) tables during decoding, leading to increased computational complexity and operational costs.

Innovation Solution

Implementing techniques for updating and resetting HMVP tables after specific intervals or regions, such as every P number of CUs or every N CTU rows, and employing pruning operations to reduce redundant information, thereby optimizing the HMVP table usage during decoding.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If the HMVP table is updated after every CU decoding, then motion prediction accuracy is improved, but computational complexity and processing time increase

Engineering Contradiction:

Improvemotion prediction accuracyVSAvoidcomputational complexity

Core Design Contradiction:

Measurement precisionVSDevice complexity

Solution Approach 1:

The HMVP table is updated periodically after every P CUs are decoded, rather than after every single CU. This periodic update strategy maintains motion prediction accuracy by ensuring the table is refreshed frequently enough, while significantly reducing computational complexity by avoiding updates after every individual CU decoding operation.

Inventive Principle:

Principle #19Periodic action

Solution Approach 2:

The patent applies pruning operations to the HMVP table, keeping only a subset of motion information entries rather than maintaining all possible entries. This partial action approach reduces the table size and associated computational burden while retaining sufficient motion prediction capability for effective video decoding.

Inventive Principle:

Principle #16Partial or excessive action

2Productivity

If the HMVP table is updated after every CU decoding, then coding efficiency is improved, but processing speed decreases

Engineering Contradiction:

Improvecoding efficiencyVSAvoidprocessing time

Core Design Contradiction:

ProductivityVSLoss of time

Solution Approach 1:

By updating the HMVP table periodically after every P CUs rather than after every CU, the patent reduces the frequency of update operations. This maintains coding efficiency through sufficient table freshness while reducing processing time by eliminating redundant update operations between periodic intervals.

Inventive Principle:

Principle #19Periodic action

Solution Approach 2:

The patent performs batch updates of the HMVP table after decoding P CUs, rather than individual updates. This preliminary batching approach allows for more efficient processing by consolidating update operations, reducing overall processing time while maintaining coding efficiency.

Inventive Principle:

Principle #10Preliminary action

3Quantity of substance

If the HMVP table is pruned to remove redundant information, then memory usage is reduced, but motion information availability may be limited

Engineering Contradiction:

Improvememory usageVSAvoidmotion information availability

Core Design Contradiction:

Quantity of substanceVSLoss of information

Solution Approach 1:

The patent changes the parameter of HMVP table size by applying pruning operations that retain only essential motion information entries. This parameter adjustment reduces memory usage by eliminating redundant entries while carefully preserving sufficient motion information availability to maintain effective video decoding performance.

Inventive Principle:

Principle #35Parameter changes

Data Source

PatentUS12563220B2Inherited motion information for decoding a current coding unit in a video coding system

Publication Date: 2026.02.24 HFI INNOVATION INC

AI summary

A method of video decoding at a decoder can include receiving a bitstream including encoded data of a picture, decoding a plurality of coding units (CUs) in the picture based on motion information stored in a history-based motion vector prediction (HMVP) table, resetting the HMVP table for every N CTU rows. The method also provides updating the HMVP table with the motion information of the last Q of the plurality of CUs.