Cross-Component Band Offset Filtering for Chroma Reconstruction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing video coding technologies face challenges in efficiently reducing reconstruction errors in video data while maintaining acceptable video quality, particularly in lossy compression scenarios where color components are not fully retained.
Innovation Solution
Implementing cross-component offset filtering methods that use reconstructed samples of a first color component to derive an offset value for a second color component, independent of any associated gradient, to adjust the reconstruction values of the current sample, utilizing a band offset only mode for improved image quality.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If lossy compression is used to reduce bandwidth and storage requirements, then compression ratio is improved, but video quality deteriorates due to loss of original video information
Solution Approach 1:
The patent implements in-loop filtering that uses the reconstructed video signal itself as feedback to generate offset values for color components. The filter processes the reconstructed samples and applies corrections iteratively, using the reconstructed signal quality to drive further refinement of color accuracy, thereby continuously improving video quality within the lossy compression framework
Solution Approach 2:
The patent changes the parameter approach by deriving offset values from quantized versions of luma sample gradients rather than using original unquantized values. This parameter transformation allows the system to work with compressed data while still generating effective color component corrections, maintaining video quality despite lossy compression
2Manufacturing precision
If cross-component offset filtering is applied to reduce reconstruction errors, then video quality is improved, but device complexity increases
Solution Approach 1:
The patent segments the filtering process into distinct functional modules: gradient calculation for luma samples, quantization of gradients, classification of chroma samples based on quantized gradients, and offset application. This segmentation allows each module to be optimized independently and simplifies the overall implementation complexity while maintaining effective cross-component offset filtering
Solution Approach 2:
The patent introduces quantized luma gradient values as an intermediary between the original luma samples and the chroma sample offset correction. This intermediary representation simplifies the relationship between luma and chroma components, enabling efficient cross-component filtering without directly processing complex gradient relationships
3Device complexity
If band offset only mode is used to simplify processing, then device complexity is reduced, but reconstruction accuracy may be compromised
Solution Approach 1:
The patent applies parameter changes by using quantized versions of the luma gradient values instead of precise floating-point gradients. This quantization reduces computational complexity and simplifies the filtering process while the classification mechanism ensures that appropriate offset values are still selected to maintain reconstruction accuracy for different chroma sample types
Data Source
AI summary
Various implementations described herein include methods and systems for coding video. In one aspect, a video bitstream includes a current image frame and a first syntax element. The first syntax element has a first predefined value that indicates that the band offset only mode is enabled to determine a first sample offset of a first color sample based on luma samples, independently of any associated luma gradient of the luma samples. When the band offset only mode is enabled, quantized values are generated based on the luma samples including a first luma sample collocated with the first color sample. A first color sample is classified based on the quantized values to determine the first sample offset of the first color sample. The current image frame is reconstructed at least by adjusting the first color sample based on the first sample offset.