Motion Vector Bank for Video Encoding Candidate Expansion

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Conventional video compression techniques face inefficiencies in encoding motion vectors, particularly at lower data rates or higher compression ratios, due to limitations in utilizing distant blocks as candidate motion vectors, which restricts the number of available reference motion vector candidates.

Innovation Solution

The implementation of motion vector banks associated with different reference frame types allows for the storage and retrieval of motion vectors from coded blocks, enabling the identification of additional candidate motion vectors beyond the conventional scanning range, thereby expanding the candidate list when necessary.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Quantity of substance

If conventional video compression techniques are used with limited scanning range for motion vectors, then device complexity is reduced, but the number of available reference motion vector candidates is insufficient

Engineering Contradiction:
Improvenumber of available reference motion vector candidatesVSAvoidcomplexity of motion vector buffer management
Core Design Contradiction:
Quantity of substanceVSDevice complexity

Solution Approach 1:

The motion vector buffer is divided into multiple banks, where each bank stores motion vectors for a specific reference frame type. This segmentation allows the system to efficiently manage and retrieve motion vector candidates without increasing overall complexity, as each bank can be independently accessed based on the reference frame type being used.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

Motion vectors are pre-stored in the appropriate banks during the encoding of previous frames. When encoding the current frame, the system can directly retrieve pre-stored motion vectors from the banks without performing additional complex searches, thereby increasing the number of available candidates without proportionally increasing computational complexity.

Inventive Principle:
Principle #10Preliminary action

2Quantity of substance

If the scanning range for motion vectors is expanded to include distant blocks, then the number of candidate motion vectors increases, but the computational resources and processing time increase

Engineering Contradiction:
Improvenumber of candidate motion vectorsVSAvoidencoding speed
Core Design Contradiction:
Quantity of substanceVSProductivity

Solution Approach 1:

Motion vectors from distant blocks are pre-computed and stored in the motion vector banks during previous frame encodings. This allows the current frame encoding to directly access these pre-computed vectors without performing additional motion estimation computations, thereby increasing candidate availability while maintaining encoding speed.

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

Instead of recomputing motion vectors for distant blocks, the system copies previously computed motion vectors from the motion vector banks into the candidate list. This copying operation is computationally inexpensive compared to motion estimation, allowing the system to expand the candidate pool without significant productivity loss.

Inventive Principle:
Principle #26Copying

3Productivity

If motion vectors from all reference frame types are utilized, then encoding efficiency is improved, but the complexity of managing multiple reference frame types increases

Engineering Contradiction:
Improveencoding efficiencyVSAvoidcomplexity of managing multiple reference frame types
Core Design Contradiction:
ProductivityVSDevice complexity

Solution Approach 1:

The system segments the motion vector storage by creating separate banks for each reference frame type. This segmentation simplifies management by allowing the encoder to select and access only the relevant bank based on the current reference frame type being used, reducing the complexity of managing all reference frame types simultaneously while still utilizing motion vectors from all types for improved encoding efficiency.

Inventive Principle:
Principle #1Segmentation

Data Source

PatentUS20240314345A1Reference motion vector candidate bank
Publication Date: 2024.09.19 GOOGLE LLC
  • US20240314345A1 patent drawing
  • US20240314345A1 patent drawing
  • US20240314345A1 patent drawing

AI summary

A method for inter-prediction includes coding a first block of a current frame using a first motion vector (MV) and a reference frame type; storing, in at least one MV buffer, the first MV and the reference frame type; identifying MV candidates for coding a current block using the reference frame type; responsive to a determination that a cardinality of the MV candidates is less than a maximum number of MV candidates identifying the first motion vector in the at least one MV buffer, and responsive to a determination that the first MV is not included in the MV candidates, adding the first MV as an MV candidate; and selecting one of the MV candidates for coding the current block.