Intra Block Copy Illumination Compensation for Accurate Video Prediction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The current intra block copy illumination compensation mode in video encoding and decoding results in poor illumination compensation, leading to inaccurate prediction and low encoding and decoding efficiency.
Innovation Solution
A multi-model intra block copy illumination compensation mode is introduced, where N groups of linear model parameters are determined, and a target group is selected for performing linear transform on the reference block to improve prediction accuracy and efficiency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If a current intra block copy illumination compensation mode is used for prediction, then the encoding and decoding process can be simplified, but the illumination compensation effect is poor resulting in inaccurate prediction and low encoding and decoding efficiency
Solution Approach 1:
The patent segments the illumination compensation process by dividing candidate blocks into multiple groups (first group and second group) with different search ranges. The first group uses a limited search range while the second group uses a full search range, allowing the system to segment the complexity across different candidates rather than applying a single complex method to all blocks.
Solution Approach 2:
The patent dynamically selects between different illumination compensation approaches based on the prediction accuracy of candidate blocks. When a candidate block from the first group achieves satisfactory prediction accuracy, the system dynamically switches to a simpler compensation method; otherwise, it transitions to using a candidate block from the second group with full search range, adapting the complexity level to the specific prediction needs.
2Power
If a current intra block copy illumination compensation mode is used for prediction, then the computational load is reduced, but the illumination compensation effect is poor resulting in low encoding and decoding efficiency
Solution Approach 1:
The patent segments the candidate block list into multiple groups with different computational requirements. The first group requires lower computational load with limited search range, while the second group accepts higher computational load for full search range. This segmentation allows the system to distribute computational effort across multiple candidates rather than concentrating it all in a single complex operation.
Solution Approach 2:
The system dynamically adjusts the computational load applied to different candidate blocks based on their potential to achieve good prediction accuracy. When early candidates from the first group provide satisfactory results, the system dynamically reduces computational load by avoiding further complex searches. When prediction accuracy is insufficient, the system dynamically increases computational load by evaluating candidates from the second group with full search range.
Data Source
Figure 1~4A
Figure 4B~8
Figure 9~12
AI summary
Provided in the present application are a video encoding method and apparatus, a video decoding method and apparatus and a device, a system and a storage medium. The methods both comprise: when the current block is encoded/decoded, determining a prediction mode of the current block, if the prediction mode of the current block is a multi-model intra-frame block copy illumination compensation mode, determining a reference block of the current block, and determining N groups of linear model parameters, N being a positive integer greater than 1; and selecting a target group of linear model parameters from among the N groups of linear model parameters, and performing a linear change on the reference block by using the target group of linear model parameters, so as to obtain a prediction block of the current block. Therefore, an illumination compensation effect is improved, thereby improving the prediction accuracy and an encoding/decoding performance.