Dynamic Bi-Directional Merge Candidates for Image Decoding

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing image coding methods using inter prediction lack efficiency in selecting optimal combinations of prediction directions, motion vectors, and reference picture indexes, leading to suboptimal coding performance.

Innovation Solution

An image coding method that determines a maximum number of merging candidates for each block, derives multiple first candidates based on neighboring blocks, and combines them to create a second candidate for bi-directional prediction, increasing the variety of combinations and coding efficiency.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Productivity

If the number of merging candidates is increased to improve coding efficiency, then the variety of prediction combinations increases, but the device complexity and processing overhead increase

Engineering Contradiction:

Improvecoding efficiencyVSAvoidcomplexity in deriving and managing merging candidates

Core Design Contradiction:

ProductivityVSDevice complexity

Solution Approach 1:

The merging candidate derivation is segmented into two distinct groups: first merging candidates derived from spatial and temporal neighboring blocks, and second merging candidates derived through combination of first candidates. This segmentation allows systematic management of candidate complexity while ensuring comprehensive coverage of prediction options.

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

Second merging candidates are generated by combining motion vectors and reference picture indexes from different first merging candidates. This combining approach increases the variety of prediction combinations without requiring additional neighboring block analyses, thus improving coding efficiency while controlling processing overhead.

Inventive Principle:

Principle #5Merging (Combining)

2Measurement precision

If more merging candidates are derived from neighboring blocks, then the accuracy of motion prediction improves, but the processing time and computational load increase

Engineering Contradiction:

Improveaccuracy of motion predictionVSAvoidprocessing time for candidate derivation

Core Design Contradiction:

Measurement precisionVSLoss of time

Solution Approach 1:

First merging candidates are derived preliminarily from spatial and temporal neighboring blocks before the actual coding decision. This preliminary derivation establishes a foundation of high-quality motion prediction candidates that can be quickly combined to form second merging candidates, reducing the need for extensive real-time analysis during encoding.

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

Motion vectors and reference picture indexes from neighboring blocks are copied to form first merging candidates, and these are subsequently combined to form second merging candidates. This copying approach leverages already-computed motion information from neighboring blocks, avoiding redundant computation while maintaining prediction accuracy.

Inventive Principle:

Principle #26Copying

3Adaptability or versatility

If bi-directional prediction candidates are always available, then the flexibility of prediction modes increases, but the complexity of candidate management increases

Engineering Contradiction:

Improveflexibility of prediction modesVSAvoidcomplexity in managing bi-directional prediction candidates

Core Design Contradiction:

Adaptability or versatilityVSDevice complexity

Solution Approach 1:

The availability of bi-directional prediction candidates is made dynamic rather than static. Second merging candidates that enable bi-directional prediction are generated on-demand by combining first merging candidates, ensuring bi-directional prediction flexibility is available when beneficial while avoiding the overhead of pre-managing all possible bi-directional candidates.

Inventive Principle:

Principle #15Dynamics

Solution Approach 2:

The prediction mode parameters (motion vectors and reference picture indexes) are changed by combining parameters from different first merging candidates to create second merging candidates. This parameter combination approach provides flexible bi-directional prediction options without requiring separate management structures for different prediction modes.

Inventive Principle:

Principle #35Parameter changes

Data Source

PatentUS12425634B2Image decoding method and image decoding apparatus using candidate motion vectors

Publication Date: 2025.09.23 SUN PATENT TRUST

AI summary

An image coding method includes: determining a maximum number of merging candidates; deriving first merging candidates based on blocks spatially or temporally neighboring a current block; determining whether or not a total number of the first merging candidates is smaller than the maximum number; deriving, by making a combination out of the derived first merging candidates, a second merging candidate for bi-directional prediction when it is determined that the total number of the first merging candidates is smaller than the maximum number; selecting a merging candidate to be used for the coding of the current block from the first merging candidates and the second merging candidate; and coding, using the determined maximum number, an index for identifying the selected merging candidate, and attaching the coded index to the bitstream.