High-Order Intra Prediction for Non-Linear Video Textures
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current intra prediction methods struggle with low accuracy in predicting non-linear texture images due to the use of one-dimensional projection prediction schemes that rely on straight-line reference pixel paths.
Innovation Solution
Implementing a high-order intra prediction (HOIP) technology that uses non-linear projection functions, such as quadratic functions, to accurately predict pixel values in non-linear texture images, and encoding bitstream information to facilitate decoding at the decoder side.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If one-dimensional projection prediction scheme is used, then prediction accuracy is high for linear texture images, but prediction accuracy is low for non-linear texture images
Solution Approach 1:
The patent implements dynamic prediction mode selection that adapts to different texture types. The system determines whether to use linear or non-linear prediction based on the actual image characteristics, allowing the prediction method to dynamically adjust to match the texture pattern being processed, thereby resolving the contradiction between high accuracy for linear textures and adaptability to non-linear textures
Solution Approach 2:
The patent changes the prediction parameters by introducing non-linear prediction functions (such as quadratic functions) alongside traditional linear prediction. By modifying the mathematical model from simple linear projection to include higher-order polynomial functions, the system can accurately represent both linear and non-linear texture patterns, thus improving adaptability while maintaining prediction accuracy
2Measurement precision
If high order intra prediction technology is implemented, then prediction accuracy for non-linear texture images is improved, but coding complexity increases
Solution Approach 1:
The patent segments the prediction process into distinct linear and non-linear modes. By dividing the prediction task into separate处理 paths based on texture type detection, the system can apply complex non-linear prediction only when necessary, rather than always using it. This segmentation reduces overall coding complexity while maintaining high prediction accuracy for non-linear textures
Solution Approach 2:
The patent applies non-linear prediction partially - only to regions identified as non-linear textures. By using the simpler linear prediction for linear texture regions and reserving complex non-linear prediction for only the necessary non-linear regions, the system achieves high prediction accuracy where needed while minimizing the overall coding complexity burden
Data Source
Figure 1
Figure 2~3
Figure 4
AI summary
The present application relates to the technical field of video coding, and disclosed thereby are an intra prediction coding method, an intra prediction decoding method, a communication apparatus, and a storage medium. Coding and decoding of high order intra prediction on a video image frame can be performed. The method comprises: generating bitstream information of a video image frame, the bitstream information of the video image frame comprising first indication information corresponding to a target primary angle prediction mode and second indication information corresponding to a target extended angle prediction mode, and the target primary angle prediction mode and the target extended angle prediction mode being angle prediction modes used to perform high order intra prediction on pixels to be predicted in a prediction unit in the video image frame; and sending the bitstream information of the video image frame to a decoding end.