Luma-Chroma Deep Learning Filtering for Video Coding Efficiency
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing video encoding and decoding technologies face challenges in improving the filtering effect of loop filtering to enhance encoding and decoding efficiency.
Innovation Solution
A deep learning-based filtering method that utilizes luma and chroma component information to generate an input parameter for a deep learning filter, enhancing the filtering performance of luma components and improving video encoding and decoding efficiency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If conventional loop filtering is applied on the reconstructed image, then the image quality is improved, but the filtering effect is limited and encoding/decoding efficiency cannot be sufficiently enhanced
Solution Approach 1:
The patent combines luma component reconstructed images with chroma component information (including chroma reconstructed images, prediction images, and block partitioning information) to create a unified input for the deep learning filter. This merging of multiple component types enables the filter to leverage correlations between luma and chroma components, achieving superior filtering effects while maintaining encoding efficiency through integrated processing.
Solution Approach 2:
The patent transforms traditional filtering parameters by introducing chroma component information as additional input parameters to the deep learning filter. This parameter expansion allows the filter to adaptively adjust filtering strength and characteristics based on chroma content, thereby improving filtering precision without significantly increasing computational complexity.
2Manufacturing precision
If chroma component information is utilized to enhance deep learning filter performance, then filtering precision is improved, but computational complexity increases
Solution Approach 1:
The patent performs preliminary processing of chroma component information during the encoding phase, including generating chroma prediction images and block partitioning information in advance. These pre-computed chroma data are then directly utilized by the deep learning filter without requiring additional real-time computation, thereby reducing runtime computational complexity while maintaining enhanced filtering precision.
Solution Approach 2:
The deep learning filter is designed to process multiple types of input data (luma reconstructed images, chroma reconstructed images, prediction images, and block partitioning information) through a unified architecture. This multi-functional design allows the same filter to leverage diverse chroma information types without requiring separate processing paths, thus improving filtering precision while controlling computational complexity through resource sharing.
Data Source
AI summary
A deep learning-based filtering method includes obtaining a reconstructed image of luma component corresponding to an encoded image and chroma component information corresponding to the encoded image; generating an input parameter of a deep learning filter according to the reconstructed image of luma component and the chroma component information; and generating, based on the input parameter to the deep learning filter, a filtered image corresponding to the reconstructed image of luma component.