Selective Candidate Refinement for Decoder-Side Video Coding

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing video coding systems face challenges in efficiently deriving motion vectors at the decoder side, leading to high computational requirements and suboptimal coding efficiency, particularly in advanced video coding standards like HEVC.

Innovation Solution

The proposed method involves refining initial motion vectors using candidate characteristics and adaptive search techniques, such as L-shape and bi-directional templates, to improve motion vector derivation efficiency and reduce computational overhead.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If bilateral template matching is applied to refine motion vectors at the decoder side, then motion vector accuracy is improved, but computational complexity increases

Engineering Contradiction:
Improvemotion vector accuracyVSAvoidcomputational complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The patent applies preliminary action by performing motion vector refinement at the decoder side using bilateral template matching before final motion compensation. The decoder derives refined motion vectors from candidate motion vectors and reference picture data, achieving accurate motion compensation without requiring additional encoder processing.

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

The patent implements self-service by enabling the decoder to autonomously perform motion vector refinement operations. The decoder uses locally available data (candidate motion vectors, reference pictures) to derive refined motion vectors, eliminating the need for complex encoder-side processing or additional bitstream information.

Inventive Principle:
Principle #25Self-service

2Productivity

If more motion vector candidates are derived and refined, then coding efficiency is improved, but processing time and computational load increase

Engineering Contradiction:
Improvecoding efficiencyVSAvoidprocessing time
Core Design Contradiction:
ProductivityVSLoss of time

Solution Approach 1:

The patent applies segmentation by dividing the motion vector refinement process into distinct stages: deriving candidate motion vectors from multiple sources (spatial neighbors, temporal references), selecting initial candidates, performing bilateral template matching refinement, and applying refined vectors to different prediction units. This segmented approach enables selective processing and optimization at each stage.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent implements partial action by deriving and refining only the necessary number of motion vector candidates based on prediction unit characteristics. The decoder selectively applies refinement to candidates that will actually be used for motion compensation, avoiding unnecessary processing of excess candidates that would increase computational load without improving coding efficiency.

Inventive Principle:
Principle #16Partial or excessive action

Data Source

PatentEP3769529B1Candidate reorganizing with advanced control in video coding
Publication Date: 2025.09.17 MEDIATEK INC
  • EP3769529B1 patent drawingFigure 1~2
  • EP3769529B1 patent drawingFigure 3~4A
  • EP3769529B1 patent drawingFigure 4B~5

AI summary

Method and apparatus for a coding system with selectively refining an initial MV based on candidate characteristics of the initial MV are disclosed. The candidate characteristics of the initial MV may correspond to candidate position within the candidate list, association within interleaved candidate groups formed by candidates in the candidate list or restricted candidates belonging to a candidate group formed by excluding one or more spatial candidates, or one or more selected-type candidates from an original candidate group. According to another method, the search for candidate refinement is restricted to 8-connected locations around the target initial MV. In yet another method, the candidate refinement uses a reduced template of the current block and a reduced corresponding template of the reference block. In yet another method, the search for candidate refinement uses an adaptive search window.