Template-Based Local Illumination Compensation for Video Decoding
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing video coding technologies face challenges in effectively addressing local illumination variations between current and reference blocks, leading to suboptimal compression efficiency and quality.
Innovation Solution
Implementing local illumination compensation (LIC) models that utilize templates of neighboring samples and reference blocks to derive parameters for compensating samples, including non-linear terms and gradient-based adjustments, to generate compensated samples.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If traditional video coding techniques are used without local illumination compensation, then the device complexity is lower, but the compression efficiency and quality deteriorate due to inability to address local illumination variations
Solution Approach 1:
The patent applies local illumination compensation by dividing the block into sub-blocks and deriving separate LIC parameters for each sub-block. This allows different parts of the block to have different compensation characteristics, addressing local illumination variations more effectively than a single global compensation model, thereby improving compression efficiency while managing complexity through localized processing.
Solution Approach 2:
The patent segments the current block into multiple sub-blocks (e.g., 4 sub-blocks) and performs independent LIC parameter derivation for each sub-block. This segmentation enables the system to capture local illumination changes in different regions separately, improving the overall compression efficiency by adapting to local variations without requiring a single complex global model.
2Manufacturing precision
If a single LIC model is used for the entire block, then the device complexity is lower, but the manufacturing precision of illumination compensation deteriorates due to inability to capture local variations
Solution Approach 1:
The patent implements multiple LIC models for different sub-blocks within the same block, allowing each sub-block to have its own illumination compensation parameters. This ensures that local illumination variations are captured with high precision in each region, improving the overall compensation accuracy without requiring an excessively complex global model.
Solution Approach 2:
The patent dynamically selects and applies different LIC models to different sub-blocks based on local characteristics. This dynamic approach allows the system to adapt to varying illumination conditions in different regions of the block, improving compensation precision while maintaining reasonable device complexity through selective application of complexity only where needed.
3Measurement precision
If multiple LIC models are derived for different sub-blocks, then the illumination compensation precision is improved, but the computational complexity and processing time increase
Solution Approach 1:
The patent divides the block into multiple sub-blocks and performs LIC parameter derivation independently for each sub-block. This segmentation allows the computational workload to be distributed and processed in parallel, reducing the overall processing time while maintaining high compensation precision through localized parameter optimization for each sub-block.
Solution Approach 2:
The patent applies LIC processing selectively to sub-blocks where it is most beneficial, rather than uniformly to the entire block. This partial action approach reduces unnecessary computational overhead in regions where illumination variations are minimal, thereby decreasing processing time while still achieving high compensation precision in regions where it matters most.
Data Source
AI summary
Processing circuitry for video decoding receives coded information of a current block in a current picture from a coded video bitstream, the coded information is indicative of applying local illumination compensation (LIC) on the current block in the current picture. The processing circuitry derives, parameters of an LIC model, according to a first template of a current block and a second template of a reference block in a reference picture. The reference block is pointed based on a motion vector for the current block. The first template includes a subset of reconstructed neighboring samples that are above and left to the current block, and the second template includes collocated samples to the subset of the reconstructed neighboring samples. The processing circuitry applies the LIC model on the current block according to the reference block to generate compensated samples of the current block.