Video Processing with Unit-Level Neural Post-Filter Activation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current video processing technologies face challenges in efficiently activating neural-network post-filters (NNPFs) at a level lower than the picture level, leading to inefficient application and suboptimal coding quality, as existing designs only support activation at the picture level and do not account for varying content within a video unit.
Innovation Solution
The proposed method controls the activation of NNPFs at a refined level, such as slice or tile level, using a set of syntax elements within a video message unit to enable precise application of multiple NNPFs for different video units, including sequences, pictures, slices, tiles, and regions of interest, thereby improving coding quality.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If NNPF activation is controlled at picture level only, then device complexity is reduced, but coding quality deteriorates due to inability to adapt to varying content within video units
Solution Approach 1:
The picture is segmented into multiple video units (slices, tiles, or blocks), and NNPF activation is controlled independently for each video unit through separate syntax elements. This allows selective application of NNPF to specific regions with varying content characteristics, improving coding quality while maintaining manageable complexity through modular control structure.
Solution Approach 2:
Different NNPF activation decisions are made for different video units within the same picture based on local content characteristics. The syntax elements enable fine-grained control where each video unit can have its own activation flag, allowing the system to adapt filter application to local content needs rather than applying a uniform picture-level decision.
2Manufacturing precision
If multiple NNPFs are applied to different video units, then coding quality is improved, but device complexity increases due to refined activation control requirements
Solution Approach 1:
The control mechanism is segmented into multiple independent syntax elements, each corresponding to a specific video unit. This segmentation allows the system to manage complexity by treating each video unit's NNPF activation as an independent control decision rather than a monolithic picture-level decision.
Solution Approach 2:
The system applies NNPF selectively to only those video units that benefit from it, rather than applying it uniformly across the entire picture. This partial action approach improves coding quality by adapting to local content needs while avoiding the excessive complexity of managing multiple filter configurations for all possible regions.
3Manufacturing precision
If NNPF is applied uniformly across the entire picture, then ease of operation is maintained, but coding quality deteriorates due to lack of content adaptation
Solution Approach 1:
The uniform picture-level control is segmented into multiple video unit-level controls. Each video unit has its own syntax element for NNPF activation, enabling the system to maintain operational simplicity through a standardized control mechanism while achieving content-adapted coding quality through selective application.
Solution Approach 2:
The NNPF activation control transitions from a static picture-level decision to a dynamic video unit-level decision process. The syntax elements enable the system to adaptively determine NNPF application for each video unit based on content characteristics, making the control mechanism dynamic rather than static.
Data Source
AI summary
Embodiments of the present disclosure provide a solution for video processing. A method for video processing is proposed. The method comprises: performing a conversion between a current video unit of a video and a bitstream of the video, wherein the bitstream comprises at least one set of syntax elements for activating at least one neural-network post-filter (NNPF) for the current video unit, and the current video unit is a portion of a current picture of the video.