Neural Network Post-Filter Group SEI Message Handling
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image encoding and decoding technologies face inefficiencies in handling intermediary pictures output by neural-network post-filter groups in SEI messages, leading to increased transmission and storage costs for high-resolution and high-quality images.
Innovation Solution
An image encoding/decoding method that includes obtaining neural network post-filter group characteristics SEI messages, generating a list of output images, and setting indicators to determine which images are included in the final output, allowing for efficient transmission and storage of only necessary images.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If all intermediary pictures output by neural-network post-filter groups are transmitted and stored, then complete image processing results are achieved, but transmission cost and storage cost increase
Solution Approach 1:
The patent extracts only the necessary final output images from the complete list of intermediary pictures generated by neural-network post-filter groups. By identifying and selecting only the required images for transmission and storage, the system removes unnecessary data while preserving the essential processing results, thereby reducing transmission and storage costs without compromising processing completeness.
Solution Approach 2:
The patent segments the complete list of intermediary pictures into two categories: necessary final output images and unnecessary intermediary pictures. This segmentation allows the system to handle different image sets differently - transmitting and storing only the necessary images while discarding or ignoring the unnecessary ones, thus resolving the contradiction between completeness and data volume.
2Manufacturing precision
If neural-network post-filter groups process all images, then high-quality output is achieved, but processing time and computational resources increase
Solution Approach 1:
The patent applies partial action by processing images through neural-network post-filter groups only to the extent necessary for producing the required final output images. Instead of uniformly processing all possible intermediary pictures, the system performs processing selectively - applying the full neural-network post-filter processing only where needed to generate necessary output images, thereby reducing overall processing time and computational resource consumption while maintaining high image quality for the required outputs.
3Quantity of substance
If intermediary pictures are not properly handled, then transmission and storage costs are reduced, but undesired output images are produced
Solution Approach 1:
The patent implements feedback mechanisms through indicator settings that provide information about which intermediary pictures are necessary and which are not. The system uses these indicators to control the transmission and storage of intermediary pictures, ensuring that necessary images are preserved while unnecessary ones are discarded. This feedback-driven approach maintains output image quality by ensuring that only appropriate images are transmitted and stored, preventing the production of undesired output images.
Data Source
AI summary
An image decoding method including obtaining a neural network post-filter (NNPF) group (NNPFG) characteristics (NNPFGC) supplemental enhancement information (SEI) message specifying information related to a group of two or more NNPFs to be applied to image data; obtaining a first list of output images that are output by one or more of the two or more NNPFs; obtaining a number of output images of the first list that are to be included in a final output of images; based on the obtained number being greater than zero, setting one or more indicators for the output images indicating whether each output image is to be included in the final output of images; and based on the first list and the one or more indicators, generating a second list of images that are to be output in the final output of images.