Image Decoding Geometric Block Partitioning for Object Boundaries

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Conventional image decoding techniques struggle with selecting appropriate block partitioning for object boundaries that are not aligned with rectangular block boundaries, leading to increased prediction errors and reduced subjective image quality.

Innovation Solution

Implementing geometric block partitioning merge to target blocks divided in a rectangular shape, which includes a merge unit that specifies and applies geometric block partitioning patterns to improve block partitioning alignment with object boundaries.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Device complexity

If rectangular block partitioning is used for all blocks, then the device complexity is low and processing is simple, but the coding performance deteriorates when object boundaries are not aligned with rectangular block boundaries

Engineering Contradiction:

Improveblock partitioning complexityVSAvoidcoding performance

Core Design Contradiction:

Device complexityVSManufacturing precision

Solution Approach 1:

The patent divides the target block into multiple sub-blocks based on geometric patterns (triangular, quadrangular, pentagonal, hexagonal) that can better align with object boundaries. This segmentation allows the partitioning to adapt to arbitrary object boundary directions while maintaining manageable complexity through standardized geometric templates.

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The patent introduces dynamic block partitioning where the partitioning shape and orientation can be selected based on the detected object boundary characteristics. The system dynamically adjusts the partitioning geometry (using flags like geo_partition_mode) to match the object boundary, rather than using fixed rectangular partitions for all cases.

Inventive Principle:

Principle #15Dynamics

2Ease of operation

If rectangular block partitioning is used, then the ease of operation is high and implementation is simple, but the prediction accuracy deteriorates for object boundaries in arbitrary directions

Engineering Contradiction:

Improvepartitioning implementation easeVSAvoidprediction accuracy

Core Design Contradiction:

Ease of operationVSMeasurement precision

Solution Approach 1:

The patent changes the partitioning parameters from fixed rectangular shapes to variable geometric shapes (triangular, quadrangular, pentagonal, hexagonal) with adjustable orientations. The system uses parameters like geo_partition_mode and rotation angles to control the partitioning geometry, allowing adaptation to object boundary directions while maintaining systematic control.

Inventive Principle:

Principle #35Parameter changes

Solution Approach 2:

The patent employs asymmetric geometric partitioning patterns that can rotate and orient themselves to match object boundaries in arbitrary directions. Instead of symmetric rectangular partitions, the system uses geometric shapes that can be rotated to create asymmetric partitions better suited to the local object boundary orientation.

Inventive Principle:

Principle #4Asymmetry

3Manufacturing precision

If geometric block partitioning merge is applied to target blocks, then the coding performance improves by reducing prediction errors, but the device complexity increases

Engineering Contradiction:

Improvecoding performanceVSAvoidpartitioning system complexity

Core Design Contradiction:

Manufacturing precisionVSDevice complexity

Solution Approach 1:

The patent applies geometric block partitioning merge selectively to specific target blocks based on their characteristics (such as object boundary presence). Not all blocks undergo geometric partitioning - only those where it benefits coding performance. This local application reduces the overall complexity increase while maintaining performance improvements where needed.

Inventive Principle:

Principle #3Local quality

Solution Approach 2:

The patent implements geometric block partitioning merge as an optional enhancement rather than a mandatory process for all blocks. The system can choose to apply partial geometric partitioning (using flags to enable/disable) or full geometric partitioning depending on the specific case, allowing flexibility in managing complexity versus performance trade-offs.

Inventive Principle:

Principle #16Partial or excessive action

Data Source

PatentUS12615359B2Image decoding device, image decoding method, and program

Publication Date: 2026.04.28 KDDI CORP

AI summary

An image decoding device includes a merge unit configured to apply geometric block partitioning merge to a target block divided in a rectangular shape, wherein the merge unit includes: a merge mode specifying unit configured to specify whether or not the geometric block partitioning merge is applied; a geometric block partitioning unit configured to specify a geometric block partitioning pattern and further perform geometric block partitioning on the target block divided in the rectangular shape by using the specified geometric block partitioning pattern; and a merge list construction unit configured to construct a merge list for the target block subjected to the geometric block partitioning and decode motion information.