Adaptive Loop Filter Boundary Gradient Handling
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In video coding, adaptive loop filtering (ALF) faces challenges when performing calculations for samples near boundaries of a picture, slice, or tile group, as neighboring samples outside these boundaries are often unavailable, leading to potential losses in subjective and objective quality due to techniques like repetitive padding.
Innovation Solution
The method involves calculating gradient values for samples near boundaries by replacing unavailable neighboring samples with the nearest available samples within the boundary, improving the accuracy and directionality of ALF, thereby enhancing the quality of filtered blocks.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If gradient calculation uses unavailable samples outside boundaries, then measurement precision of gradient values improves, but loss of information increases due to missing boundary samples
Solution Approach 1:
The patent performs preliminary actions by determining virtual gradient values for boundary samples before the actual filtering process. These virtual gradient values are calculated using available samples within the boundary, allowing the adaptive loop filter to proceed without requiring actual samples outside the boundary, thus preventing information loss while maintaining gradient accuracy for filtering decisions.
Solution Approach 2:
The patent introduces an intermediary mechanism by using available samples within the boundary to represent or substitute for unavailable boundary samples. This intermediary approach allows gradient calculations to proceed by using the nearest available samples, effectively mediating between the need for accurate gradient measurement and the absence of actual boundary samples.
2Manufacturing precision
If adaptive loop filtering is applied near boundaries, then filtering quality improves, but device complexity increases due to boundary handling requirements
Solution Approach 1:
The patent applies local quality by treating boundary samples differently from interior samples. For boundary samples, the system determines virtual gradient values using only available neighboring samples within the boundary, while interior samples use standard gradient calculations. This localized approach maintains high filtering quality at boundaries without requiring complex global boundary handling mechanisms.
Solution Approach 2:
The patent changes parameters locally at boundaries by using available samples to determine virtual gradient values instead of requiring fixed window samples. This parameter change allows the adaptive loop filter to adapt its operation at boundaries, maintaining filtering quality while avoiding the complexity of implementing separate boundary processing hardware or software modules.
3Loss of information
If gradient values are determined using available samples only, then loss of information decreases, but measurement precision of gradient directionality deteriorates
Solution Approach 1:
The patent performs preliminary determination of virtual gradient values using available samples before the adaptive loop filtering process. This preliminary action ensures that no information is lost due to missing boundary samples, while the gradient directionality is preserved through careful selection of available neighboring samples for the calculation.
Data Source
AI summary
A video encoder and/or video decoder are configured to apply an adaptive loop filter to a reconstructed block of video data. The video encoder and/or video decoder may be configured to determine a value of a gradient for a sample in the reconstructed block of video data, including performing a gradient calculation for the sample using an available sample that corresponds to an unavailable sample that is outside of one of: a picture, a slice, a tile, or a tile group that includes the reconstructed block of samples, determine an adaptive loop filter for the reconstructed block of video data based at least in part on the determined value of the gradient for the sample, and apply the determined adaptive loop filter to the reconstructed block of video data to generate a filtered block of video data.