Chroma Prediction Model Switching for Nonlinear Video Coding
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The existing video coding methods, such as VVC's CCLM, assume a linear relationship between luminance and chroma, leading to poor compression effects for coding units with non-linear correlations, resulting in increased redundancy and reduced efficiency.
Innovation Solution
Introduce a non-linear activation function into the target model to characterize the mapping relationship between luminance and chroma, allowing for flexible switching between linear and non-linear models, thereby improving chroma prediction accuracy and reducing residual values.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If a linear model (CCLM) is used to characterize the relationship between luminance and chroma, then the model complexity is low, but the prediction accuracy deteriorates for coding units with non-linear correlations
Solution Approach 1:
The patent introduces a dynamic switching mechanism that selects between linear and non-linear models based on the correlation characteristics of each coding unit. The system dynamically adjusts the prediction approach by calculating correlation coefficients and choosing the appropriate model (CCLM or non-linear) for each block, thereby adapting to varying luminance-chroma relationships without incurring unnecessary complexity for all blocks.
Solution Approach 2:
The patent changes the model parameters dynamically by introducing a switching flag and correlation coefficient thresholds. When the correlation coefficient exceeds a threshold, the system switches from the simple linear CCLM model to a more complex non-linear model, and vice versa. This parameter-based switching allows the system to optimize prediction accuracy for each coding unit based on its specific characteristics.
2Measurement precision
If a non-linear model is introduced to improve prediction accuracy for all coding units, then the prediction accuracy improves, but the computational complexity increases
Solution Approach 1:
The patent segments the coding units into two categories: those with high linear correlation (processed by simple CCLM) and those with non-linear correlation (processed by non-linear models). By segmenting based on correlation coefficient thresholds, the system applies complex non-linear processing only where necessary, rather than uniformly across all blocks, thus reducing overall computational complexity while maintaining accuracy where needed.
Solution Approach 2:
The patent applies the more computationally intensive non-linear model only partially - specifically to coding units where the linear correlation coefficient falls below the threshold. For the majority of coding units with strong linear correlation, the simpler CCLM model suffices. This partial application of the complex model avoids excessive computation while still capturing the non-linear effects where they matter.
3Ease of operation
If the CCLM model is applied universally, then the implementation is simple, but the compression efficiency deteriorates due to increased residual values
Solution Approach 1:
The patent implements a dynamic model selection mechanism that switches between CCLM and non-linear models based on correlation characteristics. This dynamic approach maintains implementation simplicity through automated threshold-based decision-making while improving compression efficiency by selecting the appropriate model for each coding unit's specific luminance-chroma relationship, thereby reducing residual values.
Solution Approach 2:
The patent incorporates feedback through correlation coefficient calculation and threshold comparison to determine model selection. The system computes the correlation coefficient for each coding unit, compares it against a threshold, and uses this feedback to select the optimal model. This feedback loop ensures that the implementation remains simple and automated while achieving improved compression efficiency through adaptive model selection.
Data Source
Figure 1~3
Figure 4~6
Figure 7~10
AI summary
A coding method, a decoding method, an apparatus and a device are provided by the present application, which are conducive to improving the applicability of the coding method and the decoding method and improving the compression effect of video coding. The decoding method of an embodiment of the present application includes: acquiring coding information corresponding to a target coding block; determining a target model according to index information of the target model; determining a chroma prediction value of the target coding block according to the target model, values of parameters and a luminance downsampling reconstruction value of the target coding block; and determining a chroma reconstruction value of the target coding block according to the chroma prediction value of the target coding block and a target difference value.