Video Coding Illumination Compensation via Adaptive Reconstructed Sample Modes
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing illumination compensation technologies in video coding are ineffective in managing illumination changes between current and reference pictures, leading to increased data redundancy and reduced compression performance.
Innovation Solution
A video coding method that performs illumination compensation by determining a reconstructed sample mode for a current block, selecting neighboring reconstructed samples from top, left, or top-and-left modes, and applying linear transformation to the prediction value based on these samples.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If existing illumination compensation technology is applied, then illumination change between current picture and reference picture is attempted to be fitted, but the compensation effect is poor and data redundancy remains high
Solution Approach 1:
The patent changes the parameters used for illumination compensation by introducing multiple candidate modes (top reconstructed sample mode, left reconstructed sample mode, top-and-left reconstructed sample mode) with different parameter selection strategies. Each mode uses different reconstructed samples from different directions to calculate linear transform parameters, allowing the system to adapt to various illumination change patterns and achieve better compensation effects while reducing data redundancy through more accurate prediction.
2Measurement precision
If multiple candidate reconstructed sample modes are introduced, then illumination compensation accuracy is improved, but coding complexity increases
Solution Approach 1:
The patent implements dynamic mode selection where the encoder chooses from multiple candidate reconstructed sample modes based on the specific characteristics of each current block. The system dynamically adapts the prediction approach by selecting the most appropriate mode (top, left, or top-and-left) for each block, improving prediction accuracy while managing complexity through conditional logic rather than uniformly applying the most complex method to all blocks.
Solution Approach 2:
The patent segments the prediction process into multiple candidate modes, each handling different spatial relationships between the current block and reference blocks. By dividing the problem into distinct modes with specific reconstruction sample selection strategies, the system can apply the most appropriate method for each scenario, balancing accuracy and complexity through structured segmentation of the prediction task.
Data Source
AI summary
A video coding method and a storage medium are provided. A determined reconstructed sample mode for a current block can be one of a top reconstructed sample mode, a left reconstructed sample mode, or a top-and-left reconstructed sample mode. As such, based on an accurate reconstructed sample mode, a neighbouring reconstructed sample of the current block and a neighbouring reconstructed sample of a reference block can be determined more accurately, then a linear transform parameter is determined based on the neighbouring reconstructed sample of the current block and the neighbouring reconstructed sample of the reference block that are determined accurately, and linear transformation is performed on a first prediction value of the current block according to the linear transform parameter to obtain an accurate second prediction value.