Illumination Compensation for Inter-Prediction Video Coding
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current video coding technologies face challenges in achieving high compression ratios with minimal sacrifice in picture quality, especially in inter-prediction coding, due to limitations in illumination compensation techniques.
Innovation Solution
The implementation of an efficient illumination compensation method for inter-prediction coding, which involves deriving parameters α and β using neighboring samples to adjust predicted samples, allowing for improved prediction accuracy and reduced computational complexity.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional illumination compensation techniques are used in inter-prediction coding, then picture quality can be maintained, but compression ratio is reduced and computational complexity increases
Solution Approach 1:
The patent extracts only the essential illumination compensation parameters (α and β) needed for effective compensation, separating them from the full set of parameters used in traditional methods. This extraction allows maintaining picture quality while reducing the data volume that needs to be transmitted, thereby improving compression ratio without sacrificing measurement precision.
Solution Approach 2:
The patent changes the parameter representation by deriving illumination compensation parameters (α and β) directly from neighboring sample values rather than transmitting full illumination models. This parameter transformation reduces the amount of data required for illumination compensation while maintaining the ability to accurately compensate for illumination variations, thus improving compression ratio while preserving picture quality.
2Measurement precision
If traditional illumination compensation techniques are used in inter-prediction coding, then picture quality can be maintained, but device complexity increases
Solution Approach 1:
The patent extracts only the critical illumination parameters (α and β) from the complex traditional illumination model, significantly reducing the computational burden. By focusing on these two key parameters derived from neighboring samples, the system maintains picture quality while reducing device complexity and computational requirements for illumination compensation.
Solution Approach 2:
Instead of using complex forward models to predict illumination variations, the patent inverts the approach by deriving parameters directly from observed neighboring sample values. This inversion simplifies the computational process by working backwards from actual data rather than forward from theoretical models, reducing device complexity while maintaining measurement precision.
3Measurement precision
If complex parameter derivation methods are used for illumination compensation, then prediction accuracy can be improved, but processing speed decreases
Solution Approach 1:
The patent extracts only the essential parameters (α and β) needed for accurate prediction, eliminating the need to compute and transmit the full set of illumination model parameters. This extraction maintains prediction accuracy by preserving the critical information while reducing processing speed requirements, as fewer parameters need to be calculated and transmitted.
Solution Approach 2:
The patent performs preliminary derivation of illumination parameters (α and β) from neighboring samples during the encoding phase, so that these parameters are ready for use during decoding. This preliminary action ensures prediction accuracy is maintained while reducing processing speed requirements during the actual decoding and reconstruction process, as the complex derivation has already been completed.
Data Source
Figure 1A
Figure 1B
Figure 2
AI summary
Embodiments provide methods and apparatus for illumination compensation for inter prediction coding of a picture, e.g. for video encoding and decoding. The methods may comprise: obtaining inter-predicted sample values for a current block of the picture; obtaining a first target value (x A ) and a second target value (x B ) of neighboring samples from a set L of neighboring samples of a reference block, wherein said set L comprises neighboring samples of at least one reference block of the current block, and obtaining corresponding positions A of the sample having the first target value and B of the sample having the second target value relative to the position of the reference block; obtaining respective values of neighboring samples (y A , y B ) of the current block at the obtained positions A and B relative to the position of the current block; obtaining values of updating parameters using the calculated values; and updating the values of the inter-predicted samples based on the updating parameters.