Mode-Dependent Transform Selection in Video Coding

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing video compression techniques are inefficient in reducing data volume for digital video streams, as they do not effectively adapt to different prediction modes, leading to suboptimal compression performance.

Innovation Solution

The implementation of mode-dependent and fixed transforms in video coding, where the transform mode is determined based on the prediction mode, using a combination of fixed transforms and mode-dependent, learned transforms to optimize compression for inter- and intra-prediction modes.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Device complexity

If fixed transforms are used for all prediction modes, then device complexity is reduced, but compression efficiency deteriorates

Engineering Contradiction:
Improvetransform processing complexityVSAvoidcompression efficiency
Core Design Contradiction:
Device complexityVSProductivity

Solution Approach 1:

The patent implements dynamic transform selection where the transform type is adapted based on the prediction mode. Different transform types (first transform type for inter-prediction, second transform type for intra-prediction) are selected according to the prediction mode, allowing the system to optimize compression efficiency for each mode while maintaining manageable complexity through structured transform sets.

Inventive Principle:
Principle #15Dynamics

2Productivity

If mode-dependent learned transforms are used, then compression efficiency is improved, but device complexity increases

Engineering Contradiction:
Improvecompression efficiencyVSAvoidtransform processing complexity
Core Design Contradiction:
ProductivityVSDevice complexity

Solution Approach 1:

The patent segments the transform processing into distinct sets: first transform types for inter-prediction modes and second transform types for intra-prediction modes. Each set contains fixed transforms and mode-dependent learned transforms. This segmentation allows the system to apply complex learned transforms only where beneficial while keeping the overall structure organized and manageable.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent applies different transform qualities locally according to prediction mode. Mode-dependent learned transforms are applied specifically to blocks with inter-prediction or intra-prediction modes where they provide compression benefits, while fixed transforms are used elsewhere. This localized application optimizes compression efficiency without unnecessarily increasing complexity across all processing paths.

Inventive Principle:
Principle #3Local quality

3Productivity

If transform adaptation to prediction modes is implemented, then compression performance is improved, but processing overhead increases

Engineering Contradiction:
Improvecompression performanceVSAvoidprocessing overhead
Core Design Contradiction:
ProductivityVSDevice complexity

Solution Approach 1:

The patent performs preliminary classification of blocks according to their prediction mode before applying transforms. The encoder and decoder determine the prediction mode (inter or intra) and select the appropriate transform type in advance, avoiding complex real-time decisions during transform processing. This preliminary action reduces processing overhead while maintaining the benefits of mode-adapted transforms.

Inventive Principle:
Principle #10Preliminary action

Data Source

PatentUS20240323361A1Combination of mode-dependent and fixed transform types in video coding
Publication Date: 2024.09.26 GOOGLE LLC
  • US20240323361A1 patent drawing
  • US20240323361A1 patent drawing
  • US20240323361A1 patent drawing

AI summary

Decoding video data includes, for a block encoded using a prediction mode, determining a transform mode for the block using the prediction mode. The transform mode is a first mode when the prediction mode is an inter-prediction mode and is a second mode when the prediction mode is an intra-prediction mode. The first mode is an available first transform type that is a combination of transforms selected from first fixed transforms and first learned transforms that each comprise a respective transformation matrix generated iteratively using blocks predicted using the inter-prediction mode. The second mode is an available second transform type that is a combination of transforms selected from second fixed transforms, which is a proper subset of the first fixed transforms, and a second learned transform comprising a transformation matrix that is generated iteratively using blocks predicted using the intra-prediction mode. Decoding the block uses the prediction and transform modes.