Neural-Network Post-Filter Input Selection Across Previous CLVS
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing video coding standards face challenges in efficiently processing and signaling input pictures for neural-network post-processing filters, particularly in versatile video coding (VVC) and versatile supplemental enhancement information (VSEI) standards, leading to inefficiencies in bandwidth usage and decoding processes.
Innovation Solution
The method involves determining the input picture for a neural-network post-filter (NNPF) within a previous coded layer video sequence (CLVS) and performing conversions between visual media data and bitstreams, utilizing flags and picture order count values to ensure proper input picture selection and updating, thereby optimizing the processing of input pictures for NNPF activation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If input pictures for neural-network post-filter are selected from previous CLVS, then decoding performance is improved, but complexity of picture selection and conversion processes increases
Solution Approach 1:
The patent applies preliminary action by pre-determining which pictures from previous CLVS can be used as input to the neural-network post-filter before the actual filtering process. The system establishes rules in advance for selecting input pictures based on their temporal relationship with the current picture, thereby streamlining the overall process while maintaining high decoding performance.
Solution Approach 2:
The patent introduces an intermediary mechanism through the use of conversion parameters and picture order count values that mediate between the visual media data and the bitstream. These intermediaries facilitate the complex picture selection and conversion processes by providing standardized reference values that simplify the determination of valid input pictures from previous CLVS.
2Productivity
If conversion between visual media data and bitstream is performed using NNPF, then processing efficiency is improved, but bandwidth requirements increase
Solution Approach 1:
The patent applies parameter changes by modifying the picture order count values and conversion parameters dynamically based on the selected input pictures from previous CLVS. These parameter adjustments optimize the conversion process between visual media data and bitstream, improving processing efficiency while controlling bandwidth requirements through adaptive parameter selection.
3Measurement precision
If picture order count values are updated for multiple CLVS, then accuracy of input picture derivation is improved, but computational complexity increases
Solution Approach 1:
The patent applies segmentation by dividing the picture order count value updates into discrete steps for each involved CLVS. Instead of processing all CLVS simultaneously, the system iteratively updates values for each CLVS starting from the second involved CLVS until the last involved CLVS, thereby maintaining high accuracy while reducing computational complexity through systematic segmentation of the update process.
Data Source
AI summary
A mechanism for processing media data is disclosed. The mechanism includes determining to allow an input picture to a neural-network post-filter (NNPP) to be in a previous coded layer video sequence (CLVS) that precedes a CLVS containing a current picture for which the NNPF is activated. A conversion is performed between a visual media data and a bitstream based on the NNPF.