Cross-Partition Frequency Transform for Video Coding Efficiency

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Conventional moving image coding technologies like H.264/AVC do not effectively utilize spatial correlation between adjacent regions with partition boundaries, limiting coding efficiency when high spatial correlation exists across partitions.

Innovation Solution

A moving image coding device that divides macro blocks into transformation target regions across partitions, applies frequency transformation to these regions, and integrates adjacent regions with high spatial correlation, while transmitting integration information to improve coding efficiency.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Productivity

If frequency transformation is applied to each partition separately in conventional H.264/AVC coding, then the coding process is simple and follows standard partition structures, but coding efficiency is limited when high spatial correlation exists across partition boundaries

Engineering Contradiction:

Improvecoding efficiencyVSAvoidtransformation region division complexity

Core Design Contradiction:

ProductivityVSDevice complexity

Solution Approach 1:

The patent segments the macro block into transformation target regions that may span multiple partitions. This segmentation allows frequency transformation to be applied to regions with high spatial correlation regardless of partition boundaries, improving coding efficiency while maintaining manageable region sizes for processing.

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The patent merges adjacent partitions into unified transformation target regions when high spatial correlation is detected across partition boundaries. This merging enables the frequency transformation to exploit spatial redundancy across partitions, directly addressing the limitation of conventional partition-based processing.

Inventive Principle:

Principle #5Merging (Combining)

2Productivity

If partition boundaries are strictly maintained for frequency transformation, then the coding structure remains simple and computationally efficient, but spatial correlation across partitions is not utilized, limiting compression performance

Engineering Contradiction:

Improvespatial correlation utilizationVSAvoidtransformation structure complexity

Core Design Contradiction:

ProductivityVSDevice complexity

Solution Approach 1:

The patent introduces dynamic adaptation where the transformation target region boundaries are not fixed by partition structures but are determined based on spatial correlation characteristics of the image content. This dynamic approach allows the transformation regions to adapt to the actual spatial relationships in the image, improving compression performance.

Inventive Principle:

Principle #15Dynamics

Solution Approach 2:

The patent applies different transformation region configurations locally based on spatial correlation characteristics. Regions with high spatial correlation across partitions are merged into larger transformation targets, while regions without such correlation maintain standard partition-based transformation, optimizing performance for each local area.

Inventive Principle:

Principle #3Local quality

3Productivity

If transformation target regions are integrated across multiple partitions, then coding efficiency improves by utilizing spatial correlation, but the amount of integration information to be transmitted increases

Engineering Contradiction:

Improvecoding efficiencyVSAvoidtransmission data amount

Core Design Contradiction:

ProductivityVSQuantity of substance

Solution Approach 1:

The patent encodes transformation target region information by specifying region boundaries and integration relationships in a compact format. Rather than transmitting full region masks, the system uses parameter-based representation that efficiently conveys the integrated region structure, minimizing the increase in transmission data.

Inventive Principle:

Principle #35Parameter changes

4Reliability

If conventional partition-based frequency transformation is used, then processing is computationally efficient and follows standard algorithms, but decoding performance is limited by inability to exploit cross-partition spatial redundancy

Engineering Contradiction:

Improvedecoding performanceVSAvoidtransformation region processing complexity

Core Design Contradiction:

ReliabilityVSDevice complexity

Solution Approach 1:

The patent performs preliminary determination of transformation target regions that span multiple partitions before applying frequency transformation. This preliminary action allows the decoding process to efficiently reconstruct high-frequency components across partition boundaries, improving decoding performance and image quality.

Inventive Principle:

Principle #10Preliminary action

Data Source

PatentUS12526408B2Moving image decoding method and moving image coding method

Publication Date: 2026.01.13 SHARP KK

AI summary

A frequency transformation determination unit determines whether a plurality of adjacent transformation target regions with the partition boundary interposed therebetween are integrated or not. A transformation coefficient generation unit generates, by applying one frequency transformation, a transformation coefficient for the transformation target regions where the frequency transformation determination unit determined to integrate.