Video Coding Illumination Compensation at Object Boundaries
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing illumination compensation methods in video encoding rely on L-shaped sets of reconstructed neighboring samples for current blocks, which may not accurately estimate IC parameters when neighboring samples belong to different objects, leading to suboptimal prediction accuracy.
Innovation Solution
Derive IC parameters jointly from sets of reconstructed samples of reference blocks, rather than using L-shaped sets of neighboring samples, particularly in bi-prediction scenarios, to improve prediction accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If L-shaped sets of reconstructed neighboring samples are used for IC parameter estimation, then the estimation process is simple and fast, but the prediction accuracy deteriorates when neighboring samples belong to different objects
Solution Approach 1:
The patent segments the reference block into multiple regions (e.g., first region and second region) and performs separate IC parameter estimations for each region. This allows the method to adapt to different object boundaries within the reference block, improving accuracy when neighboring samples belong to different objects while maintaining computational efficiency through localized processing.
Solution Approach 2:
The patent dynamically selects the IC parameter estimation method based on the characteristics of the reference block. When the reference block contains samples from different objects, the method switches to region-based segmentation; otherwise, it uses the standard L-shaped set approach. This dynamic adaptation resolves the contradiction by applying the appropriate complexity level based on actual needs.
2Reliability
If IC parameters are estimated separately for each reference block in bi-prediction, then the prediction adapts to local variations, but the computational complexity increases
Solution Approach 1:
The patent applies partial action by performing separate IC parameter estimations only when necessary (e.g., when region boundary detection indicates different objects). For uniform regions, a single estimation suffices. This partial application of the complex operation maintains reliability when needed while reducing computational complexity in simpler cases.
Solution Approach 2:
The patent applies different IC parameter estimation strategies to different local regions of the reference block. Each region undergoes estimation tailored to its specific characteristics, improving local prediction accuracy while the overall computational complexity is managed through selective application rather than uniform processing across the entire block.
3Measurement precision
If region-based segmentation is applied to reference blocks, then prediction accuracy improves at object boundaries, but the encoding complexity increases
Solution Approach 1:
The patent performs preliminary detection of region boundaries within the reference block before applying segmentation-based IC parameter estimation. This preliminary action identifies when segmentation is actually needed (at object boundaries) versus when the standard method suffices, thereby improving accuracy at critical boundaries while avoiding unnecessary encoding complexity in uniform regions.
Solution Approach 2:
The patent changes the IC parameter estimation approach based on detected region characteristics. When object boundaries are detected, the method transitions from standard L-shaped set estimation to region-based segmentation with separate parameter estimation for each region. This parameter change in the estimation strategy improves accuracy at boundaries while the conditional application limits overall encoding complexity.
Data Source
AI summary
Methods and apparatuses for video coding and decoding are provided. The method of video encoding includes determining a set of parameters for illumination compensation associated with a first motion compensated reference block of a block in a picture of a video based on a function of a set of samples of the first motion compensated reference block and a set of samples of a second motion compensated reference block of the block, processing a prediction of the block based on the set of parameters, the prediction being associated with the first motion compensated reference block and encoding the block based on the processed prediction. A bitstream formatted to include encoded data, a computer-readable storage medium and a computer program product are also described.