Convolutional Cross-Component Prediction for Video Coding Efficiency
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing video coding technologies face inefficiencies in compressing video data due to limited bandwidth and memory resources, particularly in predicting chroma samples from luma samples, leading to suboptimal coding efficiency.
Innovation Solution
Implementing cross-component prediction methods such as Edge-classified Linear Model (ELM), Filter-based Linear Model (FLM), Gradient Linear Model (GLM), Convolutional Cross-Component Model (CCCM), and Multiple Linear Regression (MLR) models to enhance the prediction of chroma samples from luma samples, utilizing various linear prediction techniques and matrix decompositions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If conventional video coding methods are used for compressing video data, then bandwidth and memory resources are limited, but coding efficiency is suboptimal
Solution Approach 1:
The patent changes the parameters of the prediction model by classifying luma samples into different groups based on edge information and applying different linear prediction models to different groups. This allows the system to adapt to different local characteristics of the video data, improving coding efficiency without requiring additional bandwidth or memory resources. The classification based on edge detection enables the system to select appropriate prediction models dynamically.
Solution Approach 2:
The patent segments the luma samples into multiple groups based on edge information characteristics. By dividing the prediction process into different segments (groups with different edge properties), the system can apply specialized prediction models to each segment, thereby improving overall coding efficiency while working within limited resource constraints.
2Reliability
If cross-component prediction is applied to predict chroma samples from luma samples, then video quality is maintained, but bit rate requirements increase
Solution Approach 1:
The patent applies different prediction models to different local regions (groups) of the video data based on their specific characteristics. By matching the prediction model to the local edge properties of each group, the system maintains high video quality in each region while optimizing the bit rate usage. This localized approach ensures that each region gets the most appropriate prediction method, reducing overall bit rate requirements.
Solution Approach 2:
The patent introduces dynamic selection of prediction models based on edge information classification. Instead of using a single static prediction model for all chroma samples, the system dynamically selects from multiple linear prediction models depending on the edge characteristics of each luma sample group. This dynamic adaptation improves video quality while efficiently managing bit rate consumption.
Data Source
AI summary
The present disclosure provides a method for decoding video data, comprising: obtaining a bitstream, obtaining an information related to one or more convolutional cross-component models (CCCMs) from the bitstream, wherein each of the one or more CCCMs comprises a filter shape and a set of weighting coefficients corresponding to the filter shape for predicting each of the chroma sample values of a video block from the bitstream based on a plurality of corresponding luma sample values, and decoding the video data based on the information related to the one or more CCCMs.