Neural Network Post-Filter Patch Size Signaling in Video Coding

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Current video coding standards, such as ITU-T H.264 and HEVC, face limitations in efficiently signaling neural network post-filter parameter information, which can impact the quality of decoded video data.

Innovation Solution

The proposed techniques involve signaling neural network post-filter characteristics messages, including syntax elements that indicate the acceptance of patch sizes for input, allowing for efficient communication of post-processing filter parameters in video coding systems, applicable to various video coding standards including future standards like VVC.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If neural network post-filter parameter information is signaled using existing video coding standards, then the implementation is simpler, but the quality of decoded video data deteriorates due to inefficient parameter communication

Engineering Contradiction:

Improvequality of decoded video dataVSAvoidcomplexity of signaling mechanism

Core Design Contradiction:

ReliabilityVSDevice complexity

Solution Approach 1:

The patent segments the neural network post-filter parameter information into distinct syntax elements within the bitstream structure. Each parameter (such as patch size dimensions, filter coefficients, and activation flags) is separately coded and transmitted, allowing the decoder to reconstruct the filter configuration accurately without requiring complex overhead signaling mechanisms.

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The patent creates a universal signaling framework that can accommodate different neural network post-filter configurations and architectures. The syntax elements are designed to be general-purpose, supporting various filter types (e.g., spatial filters, temporal filters, deep neural networks) while maintaining a consistent bitstream structure that works across different video coding applications and standards.

Inventive Principle:

Principle #6Universality (Multi-functionality)

2Reliability

If detailed neural network post-filter parameters are transmitted, then the filtering quality improves, but the bitstream size increases

Engineering Contradiction:

Improveaccuracy of post-filter applicationVSAvoidbitstream size

Core Design Contradiction:

ReliabilityVSQuantity of substance

Solution Approach 1:

The patent applies local quality by transmitting detailed neural network post-filter parameters only for specific regions or blocks where the filter is activated, rather than uniformly across the entire video sequence. The syntax elements include region-specific information such as block-level activation flags and localized patch size parameters, allowing high-quality filtering where needed while minimizing bitstream overhead in regions where filtering is not applied or uses default parameters.

Inventive Principle:

Principle #3Local quality

Data Source

PatentUS12262060B2Systems and methods for signaling neural network post-filter patch size information in video coding

Publication Date: 2025.03.25 SHARP KK

AI summary

A device may be configured to perform filtering based on information included in a neural network post-filter characteristics message. In one example, the neural network post-filter characteristics message includes a syntax element indicating whether the post-processing filter accepts as input a patch size having a width equal to horizontal sample counts indicated by a syntax element and a height equal to vertical sample counts indicated by a syntax element, or a patch size having a width equal to a multiple of horizontal sample counts indicated by a syntax element and a height equal to a multiple of vertical sample counts indicated by a syntax element.