Multi-Scale Neural In-Loop Filtering for Video Coding Efficiency
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing video coding technologies face challenges in improving coding efficiency and reducing filter complexity, particularly in video compression standards like HEVC and VVC.
Innovation Solution
Implementing a neural network-based in-loop filtering using a multi-scale neural network structure with branches for video units, tailored for luma and chroma components, and applied to intra or inter slices to enhance filtering performance and reduce complexity.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If traditional filtering methods are used in video coding, then implementation is simpler, but filtering performance is insufficient
Solution Approach 1:
The neural network filter is divided into multiple parallel branches, each processing different scale features independently. This segmentation allows the complex filtering task to be distributed across multiple simpler sub-tasks, improving overall filtering performance while managing computational complexity through parallel processing
Solution Approach 2:
The patent introduces multi-scale processing by examining video data at different resolution levels or feature scales simultaneously. This dimensional approach allows the filter to capture both fine details and broader patterns, enhancing filtering performance without proportionally increasing complexity at any single scale
2Productivity
If coding efficiency is improved through advanced filtering, then video quality increases, but computational complexity increases
Solution Approach 1:
The neural network filter applies filtering selectively rather than uniformly across all video data. By processing only certain regions, scales, or features that benefit most from neural network filtering, the system achieves improved coding efficiency while avoiding unnecessary computational overhead in areas where traditional methods suffice
Data Source
AI summary
Embodiments of the disclosure provide a solution for video processing. A method for video processing is proposed. The method includes: determining, during a conversion between a video unit of a video and a bitstream of the video, a neural network filter comprising a multi-scale neural network structure that comprises a plurality of branches; applying the neural network filter to the video unit; and performing the conversion based on the filtered video unit.