Affine Motion Derivation via Template Matching

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Current video coding techniques face challenges in efficiently compressing high-quality video data, leading to increased storage and transmission requirements, and existing affine motion modeling methods have high signaling costs and complex processing requirements.

Innovation Solution

The proposed solution involves decoder-side affine motion derivation, where affine motion parameters are determined using templates without requiring affine motion information in the bitstream, allowing for template matching based affine motion derivation to estimate complex motions like rotation and zooming without signaling affine motion parameters.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If affine motion parameters are signaled in the bitstream, then motion estimation accuracy is improved, but signaling cost and bitstream size increase

Engineering Contradiction:
Improvemotion estimation accuracyVSAvoidsignaling cost
Core Design Contradiction:
Measurement precisionVSLoss of information

Solution Approach 1:

The decoder derives affine motion parameters independently using template matching on reconstructed neighboring blocks, without requiring the encoder to signal these parameters. This self-service approach eliminates signaling overhead while maintaining accurate motion estimation through local template analysis.

Inventive Principle:
Principle #25Self-service

Solution Approach 2:

Template blocks serve as intermediaries between the encoded bitstream and the affine motion parameters. By using reconstructed neighboring blocks as templates and matching them against current block regions, the system indirectly derives accurate motion parameters without direct signaling.

Inventive Principle:
Principle #24Intermediary (Mediator)

2Adaptability or versatility

If affine motion modeling is implemented, then complex motion estimation capability is improved, but processing complexity increases

Engineering Contradiction:
Improvecomplex motion estimation capabilityVSAvoidprocessing complexity
Core Design Contradiction:
Adaptability or versatilityVSDevice complexity

Solution Approach 1:

The current block is divided into multiple template regions (e.g., top, left, and other neighboring blocks) that are processed independently. Each template region is matched separately to derive component motion vectors, which are then combined to form the final affine motion parameters. This segmentation reduces per-template processing complexity.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

Instead of performing full affine motion estimation across the entire block, the method uses partial action by applying template matching only to selected neighboring regions. This selective approach provides sufficient motion estimation capability for complex motions while significantly reducing overall processing complexity.

Inventive Principle:
Principle #16Partial or excessive action

3Productivity

If template matching based affine motion derivation is used, then coding efficiency is improved, but computational load increases

Engineering Contradiction:
Improvecoding efficiencyVSAvoidcomputational load
Core Design Contradiction:
ProductivityVSPower

Solution Approach 1:

The method uses already-reconstructed neighboring blocks as templates, which are available from previous decoding operations. This preliminary availability of template data eliminates the need for separate template generation, reducing computational load while maintaining coding efficiency through effective template matching.

Inventive Principle:
Principle #10Preliminary action

Data Source

PatentEP3596925B1Affine motion information derivation
Publication Date: 2021.03.03 QUALCOMM INC
  • EP3596925B1 patent drawingFigure 1
  • EP3596925B1 patent drawingFigure 2
  • EP3596925B1 patent drawingFigure 3

AI summary

Techniques and systems are provided for deriving one or more sets of affine motion parameters at a decoder. For example, the decoder can obtain video data from an encoded video bitstream. The video data includes at least a current picture and a reference picture. The decoder can determine a set of affine motion parameters for a current block of the current picture. The set of affine motion parameters can be used for performing motion compensation prediction for the current block. The set of affine motion parameters can be determined using a current affine template of the current block and a reference affine template of the reference picture. In some cases, an encoder can determine a set of affine motion parameters for a current block using a current affine template of the current block and a reference affine template of the reference picture, and can generate an encoded video bitstream that includes a syntax item indicating template matching based affine motion derivation mode is to be used by a decoder for the current block. The encoded video bitstream may not include any affine motion parameters for determining the set of affine motion parameters.