Visual Data Cross-Component Filtering for Better Neural Compression
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Neural network-based image and video compression technologies face challenges in achieving optimal coding quality due to the inherent difficulty in utilizing cross-component information effectively, leading to suboptimal reconstruction quality and efficiency.
Innovation Solution
A method is proposed that applies a filtering process on a second component of visual data using a neural network layer, leveraging cross-component information to enhance the quality of reconstructed visual data.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If conventional independent processing of visual data components is used, then device complexity is reduced, but coding quality deteriorates due to inability to utilize cross-component information
Solution Approach 1:
The patent merges the processing of different visual data components (e.g., luma and chroma) by using the first component as input to generate a filter kernel that is applied to the second component. This combining of component information flows enables cross-component information utilization, improving coding quality while managing complexity through a unified processing framework.
Solution Approach 2:
The patent introduces a filter kernel as an intermediary element that mediates between the first visual data component and the second component. The filter kernel is generated based on the first component and applied to the second component, serving as a bridge that enables cross-component information utilization without directly coupling the processing pipelines.
2Manufacturing precision
If cross-component information is utilized through filtering process, then coding quality is improved, but computational complexity increases
Solution Approach 1:
The patent applies local quality by generating filter kernels that are specific to local regions and component characteristics. The filtering process adapts to local variations in the visual data by using the first component's information to create localized filter kernels, improving reconstruction quality where needed while avoiding unnecessary computation in uniform regions.
Solution Approach 2:
The patent changes parameters dynamically by generating filter kernel parameters based on the first visual data component. Instead of using fixed filtering parameters, the system adapts the filter characteristics according to the content of the first component, optimizing reconstruction quality while managing computational complexity through adaptive parameter selection.
Data Source
AI summary
Embodiments of the present disclosure provide a solution for visual data processing. A method for visual data processing is proposed. The method comprises: applying, based on a first component of visual data and for a conversion between the visual data and one or more bitstreams of the visual data with a neural network (NN)-based model, a filtering process on a second component of the visual data with at least one NN layer in the NN-based model, the first component being different from the second component; and performing the conversion based on a result of the applying.