Shared Chroma Neural Network Filtering in Video Coding

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing video coding techniques require separate neural network modes for different color components, leading to increased computational complexity and processing time.

Innovation Solution

Implementing a shared neural network mode for both chroma components, with a single filtering process for both color components, reducing the need for dual execution and complexity.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If separate neural network modes are used for different color components, then filtering precision for each component can be optimized, but computational complexity and processing time increase

Engineering Contradiction:
Improvefiltering precisionVSAvoidcomputational complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The patent merges the filtering processes for first and second chroma components by using a single shared neural network mode instead of separate modes. The syntax element indicates one filtering mode that is applied to both chroma components, combining what were previously separate operations into a unified process that reduces computational complexity while maintaining filtering effectiveness.

Inventive Principle:
Principle #5Merging (Combining)

Solution Approach 2:

The patent implements universality by designing a single neural network mode that serves multiple functions - it filters both the first chroma component and the second chroma component. This universal filtering mode is indicated by a syntax element that applies to both components, allowing one mode to perform what previously required separate specialized modes.

Inventive Principle:
Principle #6Universality (Multi-functionality)

2Measurement precision

If separate neural network modes are used for different color components, then component-specific filtering can be optimized, but processing time increases

Engineering Contradiction:
Improvecomponent-specific filteringVSAvoidprocessing time
Core Design Contradiction:
Measurement precisionVSLoss of time

Solution Approach 1:

The patent combines the filtering operations for multiple chroma components into a single unified process. By using one shared neural network mode indicated by a single syntax element, the processing time is reduced because the encoder and decoder perform filtering selection and execution fewer times compared to using separate modes for each component.

Inventive Principle:
Principle #5Merging (Combining)

Solution Approach 2:

The patent establishes the shared neural network mode selection in advance through the syntax element, which defines the filtering mode for both chroma components simultaneously. This preliminary determination avoids the need for separate mode selections and executions for each component, thereby reducing overall processing time.

Inventive Principle:
Principle #10Preliminary action

3Device complexity

If a shared neural network mode is used for both chroma components, then computational complexity decreases, but adaptability for component-specific optimization is reduced

Engineering Contradiction:
Improvecomputational complexityVSAvoidcomponent-specific adaptability
Core Design Contradiction:
Device complexityVSAdaptability or versatility

Solution Approach 1:

The patent achieves universality by creating a single neural network mode that can effectively handle both chroma components. The syntax element indicates this universal mode that serves both components, providing adaptability through a unified approach rather than requiring separate component-specific modes, thus reducing complexity while maintaining versatility.

Inventive Principle:
Principle #6Universality (Multi-functionality)

Data Source

PatentUS12598314B2Neural network based filtering process for multiple color components in video coding
Publication Date: 2026.04.07 QUALCOMM INC
  • US12598314B2 patent drawing
  • US12598314B2 patent drawing
  • US12598314B2 patent drawing

AI summary

A method of processing video data includes receiving a syntax element that defines a filtering mode for a neural network (NN) model for both a first color component and a second color component, applying an instance of the NN model, in the defined filtering mode, to a first block of the first color component to generate a first filtered block, and storing the first filtered block for a coding unit (CU).