Neural-Network Post-Filter SEI Signaling for Efficient Image Coding
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The increasing demand for high-resolution and high-quality images leads to a significant increase in transmission and storage costs due to the increased amount of information, necessitating high-efficient image compression technologies that improve encoding/decoding efficiency and manage neural-network post-processing filters effectively.
Innovation Solution
An image encoding/decoding method that includes generating and signaling neural-network post-filter (NNPF) related information through supplemental enhancement information (SEI) messages, specifying filter identification and presence, allowing for efficient signaling and repetition of NNPFC SEI messages with consistent content across decoding orders.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If high-resolution and high-quality images are transmitted, then image quality is improved, but transmission cost and storage cost increase
Solution Approach 1:
The patent applies parameter changes by dynamically adjusting neural-network post-processing filter parameters based on image content characteristics. The system modifies filter strength, kernel size, and activation thresholds to optimize the balance between image quality enhancement and data compression efficiency, thereby reducing the amount of information that needs to be transmitted while maintaining high image quality.
2Productivity
If neural-network post-processing filters are applied to enhance image quality, then encoding efficiency is improved, but signaling complexity increases
Solution Approach 1:
The patent extracts and separates the neural-network post-processing filter parameters from the main bitstream, using supplemental enhancement information (SEI) messages to carry filter configuration data independently. This extraction allows the system to signal filter properties only when necessary, reducing overall signaling complexity while maintaining encoding efficiency benefits from neural-network processing.
Solution Approach 2:
The patent implements partial action by selectively applying neural-network post-processing filters only to specific regions or types of image data where they provide the most benefit. The system uses flags and indicators to signal when filters should be applied, avoiding unnecessary processing and signaling for regions where conventional processing suffices, thus reducing overall signaling complexity.
3Measurement precision
If filter property information is signaled for each NNPFC SEI message, then filter application accuracy is improved, but bitstream size increases
Solution Approach 1:
The patent applies universality by designing a reusable filter property signaling mechanism where a single set of filter parameters can be referenced multiple times through SEI messages. The system uses filter identification information and references that allow the same filter configuration to be applied across multiple images or regions without repeating the full property description, thereby reducing bitstream size while maintaining accurate filter application.
Solution Approach 2:
The patent implements preliminary action by pre-defining and registering filter property templates and configurations before they are needed in the bitstream. The system establishes a registry of available filters and their properties, allowing subsequent SEI messages to reference pre-defined filter configurations rather than signaling all properties in detail each time, thus reducing bitstream size while ensuring accurate filter application.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
An image encoding/decoding method, a method of transmitting a bitstream and a recording medium storing a bitstream are provided. The image decoding method may comprise receiving a neural-network post-filter (NNPF) related supplemental enhancement information (SEI) message to be applied to a current picture, reconstructing NNPF related information based on the NNPF related SEI message, and applying an NNPF to the current picture based on the NNPF related information. The NNPF related SEI message may include a neural-network post-filter characteristics (NNPFC) SEI message including filter identification information and filter property present information, and the NNPFC SEI message may further include base filter present information specifying whether the NNPFC SEI message includes a base neural-network post-processing filter.