Neural Network Post-Filter Overlap Removal in Video Coding
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current video coding standards, such as ITU-T H.264 and HEVC, face limitations in efficiently signaling and utilizing neural network post-filter parameter information for improved video quality, particularly in removing artifacts and enhancing visual fidelity.
Innovation Solution
The techniques involve receiving a neural network post-filter characteristics message, parsing syntax elements to derive input tensors, and generating filtered pictures using the neural network post-filter, which removes overlap regions to enhance video encoding and decoding processes.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If neural network post-filter parameters are signaled in video bitstreams, then video quality and artifact removal are improved, but device complexity and processing overhead increase
Solution Approach 1:
The patent applies parameter changes by dynamically adjusting neural network post-filter parameters (such as filter strength, kernel size, and activation thresholds) based on video content characteristics and encoding conditions. This allows the system to optimize video quality by adapting filter parameters to different scene complexities and artifact types, while avoiding unnecessary processing in regions where filtering is not needed, thus managing device complexity effectively.
Solution Approach 2:
The patent segments the video processing into multiple stages: initial decoding, overlap region identification, selective neural network filtering application, and final output generation. By dividing the processing pipeline and applying neural network filters only to specific overlap regions rather than entire frames, the system improves video quality where needed while reducing overall processing overhead and device complexity.
2Manufacturing precision
If overlap regions are removed using neural network post-filter, then visual fidelity is enhanced, but computational resources and processing time increase
Solution Approach 1:
The patent implements partial action by applying neural network post-filters selectively only to identified overlap regions rather than processing entire video frames. The system determines which regions contain artifacts and applies filtering only to those specific areas, achieving improved visual fidelity in problematic regions while minimizing unnecessary computational expenditure and processing time in already-clean regions.
Solution Approach 2:
The patent applies preliminary action by first analyzing and identifying overlap regions containing artifacts before applying the neural network post-filter. This preliminary detection stage allows the system to prepare filter parameters and region masks in advance, enabling faster and more efficient filtering execution, thereby reducing overall processing time while maintaining high visual fidelity in the filtered regions.
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 a syntax element indicating overlapping horizontal and vertical sample counts of adjacent input tensors of a neural network post-filter corresponding to the neural network post-filter characteristics message. In one example, the device may generate a filtered picture using the neural network post-filter based on a derived input tensor and indicated overlapping horizontal and vertical sample counts.