Neural Network Video Filtering for Coding Efficiency
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional video coding techniques, such as MPEG-2, MPEG-4, ITU-T.263, ITU-T.264/AVC, HEVC, and VVC, suffer from low coding efficiency, which is undesirable for digital video applications.
Innovation Solution
The proposed method involves applying a combination of neural network-based filters adaptively during the conversion between video units and bitstreams, with specific syntax elements enabling the combined application of filters to improve coding efficiency and performance.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If conventional video coding techniques are used, then device complexity is reduced, but coding efficiency deteriorates
Solution Approach 1:
The patent combines multiple filters (deblocking filter, sample adaptive offset filter, and adaptive loop filter) into a unified neural network-based filtering system. This integration allows the system to perform multiple filtering functions simultaneously through a single neural network architecture, improving coding efficiency while managing device complexity through consolidated processing.
Solution Approach 2:
The patent replaces traditional mechanical filtering mechanisms (separate deblocking, SAO, and ALF filter operations) with a neural network-based system. The neural network learns optimal filtering parameters and operations data-driven, substituting the rigid, rule-based mechanical filtering approach with an adaptive, intelligence-based system that achieves superior coding efficiency.
2Adaptability or versatility
If multiple separate filters are applied, then filtering flexibility is improved, but processing complexity increases
Solution Approach 1:
The neural network filtering system is designed to perform multiple filtering functions universally. A single neural network architecture can adapt to perform deblocking, sample adaptive offset, and adaptive loop filtering operations by learning different filtering patterns from training data. This multi-functionality provides filtering flexibility while avoiding the complexity of implementing and coordinating multiple separate filter systems.
Solution Approach 2:
The patent introduces dynamic adaptability through neural network-based filtering where filtering parameters and operations are not fixed but learned and adapted based on the specific video content characteristics. The system dynamically adjusts filtering strength, type, and parameters according to local image features, providing flexibility without requiring complex manual configuration or coordination of multiple static filters.
3Productivity
If neural network filters are applied, then coding efficiency is improved, but computational requirements increase
Solution Approach 1:
The patent applies neural network filtering selectively rather than uniformly to all video blocks. The system determines which blocks benefit most from neural network filtering and applies it only to those regions, performing partial action where needed. This approach improves coding efficiency for critical regions while reducing overall computational requirements compared to applying neural network filtering to the entire video stream.
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: applying, during a conversion between a video unit of a video and a bitstream of the video unit, a plurality of filters in combination based on a model to the video unit; and performing the conversion based on the applying.