Intra Prediction Signal Refinement for Higher Video Coding Efficiency
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing video compression techniques face challenges in achieving high coding efficiency as the increase in image size, resolution, and frame rate leads to increased data amounts, necessitating improved methods, particularly in intra prediction processes.
Innovation Solution
A video coding method and apparatus utilizing a deep learning-based refinement model with variable and fixed coefficient networks to generate refined prediction signals, reducing the amount of residual data to be encoded.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If image size, resolution, and frame rate are increased to improve video quality, then the amount of data to be encoded increases, but coding efficiency deteriorates
Solution Approach 1:
The patent applies preliminary action by using deep learning-based refinement models to process and refine prediction signals before the actual encoding process. The refinement model takes predicted signals from conventional intra prediction and enhances them in advance, so that when encoding occurs, the residual data is already optimized, reducing the data volume that needs to be encoded and improving overall coding efficiency for high-resolution video
Solution Approach 2:
The patent introduces an intermediary approach by inserting a deep learning refinement model between the conventional intra prediction process and the encoding process. This intermediary refinement model acts as a mediator that processes prediction signals, removing redundant information and enhancing important features, thereby reducing the data burden on subsequent encoding operations while maintaining or improving video quality
2Productivity
If deep learning-based refinement model is applied to improve prediction signals, then coding efficiency improves, but device complexity increases
Solution Approach 1:
The patent applies segmentation by dividing the video coding process into distinct functional modules: conventional intra prediction module, deep learning refinement model module, and encoding module. The refinement model is further segmented into variable coefficient network and fixed coefficient network components. This modular segmentation allows each component to be optimized independently and facilitates easier integration and complexity management in the overall system
Solution Approach 2:
The patent utilizes parameter changes by implementing both variable coefficient networks and fixed coefficient networks within the refinement model. The variable coefficient network adapts its parameters based on input data characteristics, while the fixed coefficient network uses pre-computed parameters. This parameter change approach allows the system to balance between modeling accuracy and computational complexity, improving coding efficiency without proportionally increasing device complexity
Data Source
AI summary
A video coding method and an apparatus for refining predicted signals in intra prediction are disclosed. The video coding method and apparatus generate refined prediction signals approximating original video signals from predicted signals of intra prediction using a variable and fixed coefficient-based deep learning model to reduce the amount of data for residual signals, which are to be encoded.