Neural Network Post-Filter Signaling for Video Coding Quality
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 video quality and compression efficiency.
Innovation Solution
The proposed techniques involve signaling neural network post-filter characteristics and syntax elements that specify the number of interpolated pictures generated by a post-processing filter, allowing for improved video encoding and decoding processes, particularly in future video coding standards like VVC.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If neural network post-filter parameter information is not signaled in current video coding standards, then the signaling complexity remains low, but the video quality deteriorates due to inefficient utilization of neural network post-filters
Solution Approach 1:
The patent segments the signaling of neural network post-filter parameters into distinct syntax elements that can be independently controlled and transmitted. Different parameter types (e.g., filter strength, filter type, interpolation factors) are signaled separately, allowing the system to transmit only necessary information for achieving video quality improvement without overwhelming signaling complexity.
Solution Approach 2:
The patent introduces new syntax elements and parameter structures specifically designed for neural network post-filter signaling. By defining standardized parameter formats and signaling mechanisms, the patent enables efficient transmission of filter characteristics information, transforming the way video coding systems communicate post-filter parameters and improving video quality through better parameter utilization.
2Manufacturing precision
If neural network post-filter characteristics are not communicated, then the bitstream remains compact, but the decoding quality deteriorates
Solution Approach 1:
The patent implements partial signaling by introducing syntax elements that convey only the essential neural network post-filter characteristics needed for quality improvement. Rather than transmitting complete filter models or excessive parameter details, the system signals key parameters (such as filter strength indicators, filter type identifiers, and interpolation factors) that provide sufficient information to enhance decoded video quality while maintaining bitstream compactness.
3Manufacturing precision
If comprehensive neural network post-filter parameters are signaled, then the video quality improves, but the encoding complexity increases
Solution Approach 1:
The patent designs universal syntax elements and signaling structures that can accommodate multiple types of neural network post-filter parameters through a unified framework. The same signaling mechanism can convey different filter characteristics (strength, type, interpolation factors) depending on the specific application, reducing encoding complexity by avoiding the need for separate specialized signaling paths for each parameter type.
Data Source
AI summary
A device may be configured to signal neural network post-filter parameter information. In one example, a device signals a neural network post-filter characteristics message including a syntax element specifying a number of interpolated pictures generated by a post-processing filter between consecutive pictures used as input for the post-processing filter. In one example, a neural network post-filter characteristics message includes a syntax element specifying a number of decoded output pictures used as input for the post-processing filter.