Neural Network Post-Filter Signaling for Video Compression
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current video coding standards face challenges in efficiently signaling neural network post-filter parameter information for coded video, which affects the quality and efficiency of video compression.
Innovation Solution
The proposed solution involves signaling a neural network post-filter characteristics message, which includes syntax elements specifying the number of interpolated pictures generated by a post-processing filter for each sublayer associated with the message.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If neural network post-filter parameter information is signaled in video coding, then video compression quality is improved, but bitstream complexity increases
Solution Approach 1:
The patent applies parameter changes by dynamically adjusting the signaling of neural network post-filter parameters based on video content characteristics and coding conditions. The system selectively signals parameters such as filter strength, kernel size, and activation flags only when beneficial, transforming the fixed parameter approach into a dynamic one that adapts to local video features, thereby improving compression quality without uniformly increasing bitstream complexity across all scenarios
Solution Approach 2:
The patent segments the neural network post-filter parameter signaling into multiple independent components that can be selectively applied to different video regions, blocks, or prediction units. This segmentation allows the encoder to apply complex neural network filtering only where needed in the video stream, rather than uniformly across the entire frame, thus improving overall quality while distributing and reducing the total bitstream overhead through selective application
2Manufacturing precision
If post-processing filter generates more interpolated pictures, then video quality is improved, but processing time increases
Solution Approach 1:
The patent implements partial action by generating interpolated pictures only for specific regions or blocks within the video frame where motion complexity or quality requirements demand it, rather than applying interpolation uniformly across the entire frame. The system uses syntax elements to indicate which sublayers or regions receive additional interpolated pictures, applying the filtering operation partially where beneficial while skipping areas where standard processing suffices, thus improving quality in critical regions without proportionally increasing overall processing time
Solution Approach 2:
The patent applies dynamics by making the number of interpolated pictures generated by the post-processing filter adaptive rather than static. The system dynamically adjusts the interpolation intensity and number of generated pictures based on local video content characteristics, motion vectors, and coding parameters, allowing the processing complexity to vary frame-by-frame or block-by-block, thereby optimizing the balance between quality improvement and processing time consumption
Data Source
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 an instance of syntax element specifying a number of interpolated pictures generated by a post-processing filter for each of a number of sublayers associated with the neural network post-filter characteristics message.